CCCC
Open-source desktop app for Mac, Windows and Linux. Agents run on your computer or on your own server.
GitHub repository for CCCCChesterRa/ccccHomepage of CCCCcccc.sh Compare with others Missing or wrong data?
- Desktop app
- Mac, Windows and Linux
- Browser
- Yes
- Phone
- Phone browser
- Command line
- Yes
- Agents
- Claude Code, Codex, Cursor Agent, OpenCode, Pi, Grok and Antigravity
- Your subscription
- Works with your own plan
- Agents run
- On your computer and on your own server
- Laptop closed
- Keeps going
- Isolation
- Worktree
- Parallel agents
- Yes
- Scheduled runs
- Yes
- Coordinator agent
- Yes
- Built-in issues
- Yes
- Pull requests
- No
- Review screen
- Yes
- Team
- Roles
- Self-hosting
- Yes
- Licence
- Apache-2.0
- Price from
- Free
- How it charges
- No paid plan and bring your own key
- Paid plans from
- Not listed
Receipts
- Checked
- Findings
- 49 published · 49 cross-checked by a second agent · read them all
- Sources
- 40 public pages on 3 sites: github.com, cccc.sh, account.cccc.sh
- GitHub stars
- 1,309 on · github.com · repository attention, not users or capability
- Screenshot
- None accepted yet
- Data
- /products/cccc.json · .md
- Method
- Same rules for all 209 researched ADEs. How we check · Report wrong data
What it does
Where you use it
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
Source and scope for Where you use itAgents it runs
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
Source and scope for Agents it runsAccounts and subscriptions
Agents run on the user own existing subscriptions and logins: connected sessions keep their app process, account, model, approvals and terminal, and CCCC needs no provider API key. Managed agents are launched from the user-installed CLIs. No vendor credits, paid CCCC plans or card exist; per-actor private environment supports different model/API stacks per agent. The only account is the optional free CCCC account for Connect/Reach; local use needs none.
Source and scope for Accounts and subscriptionsWhere agents run
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
Source and scope for Where agents runHow work is kept apart
Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies.
Source and scope for How work is kept apartParallel work
Several agents run concurrently in a Group and several Groups run side by side; the sidebar, terminal panel and status counts show running, waiting, failed and finished work across them. No concurrency quotas are documented; observed-concurrency evidence is limited to status/running counts rather than measured parallelism.
Source and scope for Parallel workScheduled runs
Daemon-owned automation runs interval, cron and one-time rules with notify plus one-time group-state/actor-control actions, completion semantics and Run-now; delivery retries are bounded and uncertain handoffs need explicit retry confirmation. No external event triggers exist. Automation needs the daemon host running (service manager for unattended); paused/stopped groups run nothing and missed fires are not replayed.
Source and scope for Scheduled runsGit and pull requests
Git stays terminal- and agent-side: CCCC surfaces read-only diffs and conflict displays, documents branch/worktree conventions (including Claude background worktrees) and conflict inspection, but never stages, discards, commits, pushes, rebases or opens pull requests, and owns no PR state.
Source and scope for Git and pull requestsLicence and source
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
Source and scope for Licence and sourcePrice
CCCC has no paid plan and no prices: free Apache-2.0 download (install script, pip wheel, Docker, source) with no currency, intervals, seats, overages, taxes or credits anywhere in the checked sources. Agent spend is the user own existing subscriptions or API accounts, handled provider-side.
Source and scope for PriceTyped filters
Every value links to its receipt below. Read plan and surface restrictions before using a filter to decide.
| Filter | Value |
|---|---|
| Desktop OS: macOS | Yes cross-checked |
| Desktop OS: Windows | Yes cross-checked |
| Desktop OS: Linux | Yes cross-checked |
| Web client | Yes cross-checked |
| Phone access: Native iOS | Unknown |
| Phone access: Native Android | Unknown |
| Phone access: Phone browser | Yes cross-checked |
| CLI | Yes cross-checked |
| Coding agents: Claude Code | Yes cross-checked |
| Coding agents: Codex | Yes cross-checked |
| Coding agents: Cursor Agent | Yes cross-checked |
| Coding agents: Gemini CLI | Unknown |
| Coding agents: OpenCode | Yes cross-checked |
| Coding agents: Pi | Yes cross-checked |
| Coding agents: Grok | Yes cross-checked |
| Coding agents: Aider | Unknown |
| Coding agents: Antigravity | Yes cross-checked |
| Use my subscription | Yes cross-checked |
| Where agents run: On your computer | Yes cross-checked |
| Where agents run: On your own server | Yes cross-checked |
| Where agents run: In the vendor's cloud | No cross-checked |
| Workspace isolation: Git worktree | Yes cross-checked |
| Workspace isolation: Container | No cross-checked |
| Workspace isolation: VM or microVM | No cross-checked |
| Parallel runs | Yes cross-checked |
| Coordinator | Yes cross-checked |
| Native issue tracking | Yes cross-checked |
| Unattended triggers | Yes cross-checked |
| Shared prompting | Unknown |
| Team roles | Yes cross-checked |
| SSO | No cross-checked |
| Pull request workflow | No cross-checked |
| Review interface | Yes cross-checked |
| Product license: Apache-2.0 | Yes. Apache-2.0 cross-checked |
| Product license: MIT | No. Apache-2.0 cross-checked |
| Product license: GPL-3.0 | No. Apache-2.0 cross-checked |
| Product license: GPL-3.0-or-later | No. Apache-2.0 cross-checked |
| Product license: AGPL-3.0 | No. Apache-2.0 cross-checked |
| Product license: AGPL-3.0-or-later | No. Apache-2.0 cross-checked |
| Product license: Elastic License 2.0 | No. Apache-2.0 cross-checked |
| Product license: Custom terms | No. Apache-2.0 cross-checked |
| Product license: Proprietary | No. Apache-2.0 cross-checked |
| Self-hosting | Yes cross-checked |
| Telemetry opt-out | Unknown |
| Free access | Yes cross-checked |
| Starting price: Free | Yes. Free cross-checked |
| Starting price: Up to $20/month | No. Free cross-checked |
| Starting price: $21–40/month | No. Free cross-checked |
| Starting price: $41–100/month | No. Free cross-checked |
| Starting price: Over $100/month | No. Free cross-checked |
| Starting price: Contact for pricing | No. Free cross-checked |
| Pricing model: Free, no paid plan | Yes cross-checked |
| Pricing model: Free plan | Unknown |
| Pricing model: Subscription | Unknown |
| Pricing model: Per seat | Unknown |
| Pricing model: Usage-based | Unknown |
| Pricing model: Per machine or compute | Unknown |
| Pricing model: Bring your own key | Yes cross-checked |
| Pricing model: One-time purchase | Unknown |
| Paid plans from: No paid plan | Unknown |
| Paid plans from: $10 or less/month | Unknown |
| Paid plans from: $11–20/month | Unknown |
| Paid plans from: $21–40/month | Unknown |
| Paid plans from: $41–100/month | Unknown |
| Paid plans from: Over $100/month | Unknown |
| Paid plans from: Contact sales | Unknown |
| Lifecycle status: Active | Unknown |
| Lifecycle status: Sunsetting | Unknown |
| Keeps going with laptop closed | Yes cross-checked |
| Open-source license | Yes. Apache-2.0 cross-checked |
| Any desktop app | Yes cross-checked |
| Paid plan up to $20/month | Unknown |
| Paid plan up to $40/month | Unknown |
| Any phone client | Yes cross-checked |
Filter receipts
Desktop OS: macOS: Yes cross-checked
Desktop OS: macOS
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Cli: Product-owned cccc CLI covers daemon lifecycle, groups, actors, messaging, inbox, ledger search, IM, space, setup, update, web and mcp, plus an interactive terminal interface with status panel and Waiting-for-you rows.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
Desktop OS: Windows: Yes cross-checked
Desktop OS: Windows
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Cli: Product-owned cccc CLI covers daemon lifecycle, groups, actors, messaging, inbox, ledger search, IM, space, setup, update, web and mcp, plus an interactive terminal interface with status panel and Waiting-for-you rows.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
Desktop OS: Linux: Yes cross-checked
Desktop OS: Linux
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Cli: Product-owned cccc CLI covers daemon lifecycle, groups, actors, messaging, inbox, ledger search, IM, space, setup, update, web and mcp, plus an interactive terminal interface with status panel and Waiting-for-you rows.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
Web client: Yes cross-checked
Web client
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Mobile apps: No native iOS/Android app ships. The Web UI is responsive with phone layouts and a PWA manifest (standalone display); iPhone/iPad use is via browser or Home Screen web app. Phone Files view is read-only.
- Browser: The Web UI is the primary client at 127.0.0.1:8848/ui, usable over LAN/tunnel/Tailscale/Reach with an Admin token; read-only exhibit mode exists for sharing.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Cli: Product-owned cccc CLI covers daemon lifecycle, groups, actors, messaging, inbox, ledger search, IM, space, setup, update, web and mcp, plus an interactive terminal interface with status panel and Waiting-for-you rows.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- Task actions: Start, steer, approve, review and resume run from Web (run controls, in-place answers, diffs), the terminal panel (group/actor pages), the CLI (group/actor/send/reply/deliver) and IM (/send /status /pause /resume).
- Notifications: Web Push per browser (message addressed to user + open requests, bounded retries), IM bridge updates, automation notify rules with Mail/reply notices, terminal panel Waiting-for-you counts.
- Cross device continuity: The same Group is reachable from another device via LAN/tunnel/Tailscale/Reach remote access, IM bridges and Connect; drafts and reading places stay per browser (unsent drafts live only in the open page).
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
Phone access: Phone browser: Yes cross-checked
Phone access: Phone browser
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile apps: No native iOS/Android app ships. The Web UI is responsive with phone layouts and a PWA manifest (standalone display); iPhone/iPad use is via browser or Home Screen web app. Phone Files view is read-only.
- Browser: The Web UI is the primary client at 127.0.0.1:8848/ui, usable over LAN/tunnel/Tailscale/Reach with an Admin token; read-only exhibit mode exists for sharing.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- Task actions: Start, steer, approve, review and resume run from Web (run controls, in-place answers, diffs), the terminal panel (group/actor pages), the CLI (group/actor/send/reply/deliver) and IM (/send /status /pause /resume).
- Notifications: Web Push per browser (message addressed to user + open requests, bounded retries), IM bridge updates, automation notify rules with Mail/reply notices, terminal panel Waiting-for-you counts.
- Cross device continuity: The same Group is reachable from another device via LAN/tunnel/Tailscale/Reach remote access, IM bridges and Connect; drafts and reading places stay per browser (unsent drafts live only in the open page).
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
CLI: Yes cross-checked
CLI
Typed value: Yes.
A complete product-owned CLI plus MCP surfaces (daemon tools over stdio for agents; Python/TypeScript/Rust SDK clients for integrators; documented Web API endpoints), but no generic public REST API with stability promises and no generic outgoing webhooks. Create/inspect/steer/stop/review run through CLI, MCP, Web and IM; auth is loopback-local principal plus Admin Access Tokens for non-local use.
- Cli: Full cccc CLI: daemon/groups/actors/messaging/inbox/ledger/IM/space/setup/update/web/mcp with JSON doctor output and documented env vars; the terminal panel adds an interactive command menu.
- Mcp server: cccc mcp serves collaboration tools over stdio; agents see a compact core (bootstrap/help/inbox/messaging/files/context/coordination/task/connect) with capability-managed expansion packs.
- Webhooks: Programmable access, webhooks: bounded search found no generic outgoing webhooks; provider-side Webhook Routines (Grok Bot) and ChatGPT event tasks are runtime-specific inbound mechanisms.
- Operations: Create/inspect/steer/stop/review via CLI (groups/actor/send/reply/deliver/inbox/tail/ledger search), MCP (group/actor/messaging/automation/terminal/debug), Web run controls and IM commands.
- No versioned public REST API or generic webhooks; automation-capable surface is CLI/MCP/Web/IM.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
Coding agents: Claude Code: Yes cross-checked
Coding agents: Claude Code
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: Codex: Yes cross-checked
Coding agents: Codex
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: Cursor Agent: Yes cross-checked
Coding agents: Cursor Agent
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: OpenCode: Yes cross-checked
Coding agents: OpenCode
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: Pi: Yes cross-checked
Coding agents: Pi
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: Grok: Yes cross-checked
Coding agents: Grok
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Coding agents: Antigravity: Yes cross-checked
Coding agents: Antigravity
Typed value: Yes.
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Harnesses: 21 named runtimes plus custom any-command fallback; native terminal, official ACP, structured RPC/headless and browser-delivered surfaces per runtime; runtime list and availability via cccc runtime list and cccc doctor.
- Model providers: Model selection is inherited from each runtime (explicit model choices applied at start/resume; no CCCC model catalog or guessing). actor --model-label is a human-declared label CCCC does not check or use to launch. ChatGPT Web uses GPT-5.x chats via the bound conversation.
- Custom adapters: custom runtime runs any command; capabilities install from explicit local path, URL or GitHub target with recorded provenance; slash skills dispatch skill: packs.
- Session integration: Managed sessions get an agent-scoped MCP server (preflight verified before launch); native user-run Codex/Claude/Pi/OpenCode sessions integrate via MCP plus hooks/plugin/extension; Web models use a remote MCP connector.
- Mixed harness operation: Each agent chooses its own runtime and both managed and connected sessions work in the same Group; workflows document Claude + Codex + Kimi style mixes with a reviewer runtime for diversity.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
Use my subscription: Yes cross-checked
Use my subscription
Typed value: Yes.
Agents run on the user own existing subscriptions and logins: connected sessions keep their app process, account, model, approvals and terminal, and CCCC needs no provider API key. Managed agents are launched from the user-installed CLIs. No vendor credits, paid CCCC plans or card exist; per-actor private environment supports different model/API stacks per agent. The only account is the optional free CCCC account for Connect/Reach; local use needs none.
- Agent subscriptions: Bring your own agent subscription: managed agents launch from installed, signed-in CLIs; connected sessions keep their own conversation, account, model and approvals; ChatGPT/Grok voice reuse their CLI logins with no extra API key.
- Vendor credits: Subscriptions, vendor credits: bounded search of the site, docs, account pages and CLI reference found no vendor credits; automation notices may use model credits via the user own runtime logins.
- Provider restrictions: Documented runtime restrictions: Antigravity cannot join as a connected session, Web models need public HTTPS + pairing, DeepSeek is developer preview, managed agents may bypass approvals (review commands first), test boundaries Linux/WSL only.
- No CCCC-side billing or credits; all inference spend stays on the user runtime accounts.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
Where agents run: On your computer: Yes cross-checked
Where agents run: On your computer
Typed value: Yes.
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
- Local: Daemon + agents run on the local machine with state in CCCC_HOME; localhost use needs no token.
- Byo remote: LAN binding, Cloudflare tunnel, Tailscale and Reach-published own host give remote/phone access to the same instance.
- Managed cloud: Execution, managed_cloud: bounded inspection found no vendor cloud compute; Remote Access serves the user own machine and Managed Access is a tunnel, not hosted execution.
- Client disconnect: Closing the browser/IM never stops the daemon; reconnect resumes views, cursors and retained terminals from ledger authority.
- Other device continuation: Phone/browser/IM/Connect continue the same Groups from other devices; in-flight page drafts stay per browser.
- No vendor cloud; regions/resources/payer are user-side and unstated.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
Where agents run: On your own server: Yes cross-checked
Where agents run: On your own server
Typed value: Yes.
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
- Self hosted: Headless/systemd/container deployments documented with backup/restore runbook; Docker image ships agent CLIs.
- No vendor cloud; regions/resources/payer are user-side and unstated.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
Where agents run: In the vendor's cloud: No cross-checked
Where agents run: In the vendor's cloud
Typed value: No.
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
- Local: Daemon + agents run on the local machine with state in CCCC_HOME; localhost use needs no token.
- Byo remote: LAN binding, Cloudflare tunnel, Tailscale and Reach-published own host give remote/phone access to the same instance.
- Managed cloud: Execution, managed_cloud: bounded inspection found no vendor cloud compute; Remote Access serves the user own machine and Managed Access is a tunnel, not hosted execution.
- Hybrid: Connect mixes instances across machines/teams with per-instance state; compute always stays on member machines, never pooled.
- Client disconnect: Closing the browser/IM never stops the daemon; reconnect resumes views, cursors and retained terminals from ledger authority.
- Other device continuation: Phone/browser/IM/Connect continue the same Groups from other devices; in-flight page drafts stay per browser.
- No vendor cloud; regions/resources/payer are user-side and unstated.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
Workspace isolation: Git worktree: Yes cross-checked
Workspace isolation: Git worktree
Typed value: Yes.
Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies.
- Checkout: Groups attach project scopes as shared checkouts; Files/Changes/agents operate on the same working copy unless a runtime isolates.
- Branch: Branch work happens through runtime CLIs, not product operations; worktree/branch identity guidance is documented for handoffs.
- Worktree: CCCC creates and cleans up no worktrees; managed Claude Code defaults to background worktrees (configurable off), agents may create their own with manual cleanup guidance.
- Vm: Isolation, vm: bounded inspection found no VM-based execution or isolation.
- Os sandbox: Isolation, os_sandbox: no OS sandbox exists; the allowlist guide states it is not a sandbox for agent processes.
- Process: Isolation, process: agent processes are user-owned with daemon lifecycle supervision only; no process boundary is documented or located in code.
- Code separation: Per-scope workspaces separate Groups; a shared checkout has no per-agent separation unless the runtime isolates.
- Security boundary: Isolation, security_boundary: no hostile-code containment is claimed; do not expose the high-privilege Web UI or share runtime homes with untrusted users.
- Shared-checkout default; isolation is scoping, not sandboxing.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
Workspace isolation: Container: No cross-checked
Workspace isolation: Container
Typed value: No.
Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies.
- Container: Docker deployment image exists for self-hosting; per-task containers do not exist.
- Vm: Isolation, vm: bounded inspection found no VM-based execution or isolation.
- Os sandbox: Isolation, os_sandbox: no OS sandbox exists; the allowlist guide states it is not a sandbox for agent processes.
- Process: Isolation, process: agent processes are user-owned with daemon lifecycle supervision only; no process boundary is documented or located in code.
- Security boundary: Isolation, security_boundary: no hostile-code containment is claimed; do not expose the high-privilege Web UI or share runtime homes with untrusted users.
- Shared-checkout default; isolation is scoping, not sandboxing.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
Workspace isolation: VM or microVM: No cross-checked
Workspace isolation: VM or microVM
Typed value: No.
Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies.
- Vm: Isolation, vm: bounded inspection found no VM-based execution or isolation.
- Os sandbox: Isolation, os_sandbox: no OS sandbox exists; the allowlist guide states it is not a sandbox for agent processes.
- Process: Isolation, process: agent processes are user-owned with daemon lifecycle supervision only; no process boundary is documented or located in code.
- Security boundary: Isolation, security_boundary: no hostile-code containment is claimed; do not expose the high-privilege Web UI or share runtime homes with untrusted users.
- Shared-checkout default; isolation is scoping, not sandboxing.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
Parallel runs: Yes cross-checked
Parallel runs
Typed value: Yes.
Several agents run concurrently in a Group and several Groups run side by side; the sidebar, terminal panel and status counts show running, waiting, failed and finished work across them. No concurrency quotas are documented; observed-concurrency evidence is limited to status/running counts rather than measured parallelism.
- Concurrent sessions: Multiple agents per Group plus multiple Groups; terminal view pages up to four agents; start/stop per agent or per Group.
- Concurrent tasks: Planned/Active/Done tasks coexist with several agents working in parallel; tracked-send fans work out with owners.
- Quotas: Parallelism, quotas: bounded search of docs, CLI reference and daemon sources found no documented concurrency quotas or limits.
- Observed concurrency: RUNNING counts, status dots, activity rings and progress bubbles show live work; no measured concurrency figures are published.
- Status visibility: Sidebar Group rows with status and Waiting-for-you counts, terminal panel status dots, Group pages with task-state counts, /status and cccc groups/actor list.
- No quotas; concurrency visibility is status-level, not measured.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
Coordinator: Yes cross-checked
Coordinator
Typed value: Yes.
The foreman coordinates peers in a daemon-owned Group: handoffs via direct agent-to-agent messages with delivery/read/reply facts, per-agent runtime/model-label/private-env configuration, mixed-harness teams, live activity plus retained terminals, and retained results in ledger/tasks/Agent State. Follow-up runs through replies, retries, new sessions and restarts across CLI, MCP, Web and IM.
- Roles: First enabled agent is foreman (coordinates + works, manages any actor); later agents are peers (manage only self); user sits above both.
- Messages: Agent-to-agent sends with @ addressing, quote/reply threading, Mail inbox, cross-Group and Connect addressing with receipts.
- Results return: Send+Reply obligations, outcomes on tracked tasks, Agent State reports and review tables return results to requesters.
- Orchestration owner: The CCCC daemon owns orchestration: single writer for ledger/tasks/lifecycle with the foreman as product-designated coordinator role.
- Live progress: Activity snapshots, runtime dock rings/bubbles, terminal views, status dots and Waiting-for-you counts during work.
- Retained results: Ledger history, task outcomes/checklists, Agent State reports, pins and Presentation slots retain results after completion.
- Followup: Reply/forward, deliver retries, cancel-reply, actor new-session/restart, Run-now automation and re-ingest continue finished work.
- Invocation: Delegate from Web composer, CLI send/tracked-send, IM /send, MCP message/task tools and automation actions.
- Effort/turn budgets are runtime-side; auth partly inherits runtimes.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
Native issue tracking: Yes cross-checked
Native issue tracking
Typed value: Yes.
A complete native task system: create/edit/discuss/prioritize/assign/search/filter/archive/reopen in the board, idea parking in the project BACKLOG.md, explicit dispatch via tracked-send, durable Planned/Active/Done states and task/session separation with per-task metadata. No labels/tags, no epic views, no external tracker: nothing external is needed or integrated.
- Create: Tasks created in the board editor, from backlog topics (Create task), or linked via tracked-send; backlog topics and ideas created in the file view.
- Edit: Title/status/assignee/priority/outcome/checklist/notes/subtasks edit with version-conflict guards; backlog entries edit in place with file-hash fencing.
- Discuss: Messages with @mentions, reply/quote/forward, review tables and outcome discussion carry task discussion; Waiting for you collects open questions.
- Prioritize: Per-task priority field plus attention filters (blocked/waiting/handoff/unassigned) and Waiting-for-you ordering surface what matters.
- Label: Native issues, label: bounded inspection of the board editor, task ops and CLI found priority/assignee/attention but no label/tag field.
- Assign: Assignee per task, @foreman/@peers addressing, tracked-send --assignee/--waiting-on/--handoff-to delegation fields.
- Search: Board search with match quotes plus daemon ledger search (-q text, sender/recipient/kind/time filters) across history.
- Filter: Status columns, attention filters, sender filters, assignee filters and message tabs (Waiting/All/Pinned/User-related).
- Archive: Archive for finished tasks (Restore returns prior status); backlog Closed section for finished topics with Reopen.
- Reopen: Restore archived tasks, Reopen closed backlog topics, resume stopped groups and restart agents on new or resumed sessions.
- Project epic views: Board columns plus per-group sidebar status cover projects; no epic rollup or cross-group portfolio view exists.
- Idea parking: BACKLOG.md holds long-lived topics/ideas beside Tasks with entry-to-task links and outcome write-back; tasks work without one.
- Task session separation: Tasks (durable records) are distinct from agents/sessions (runtimes); several agents can attach to work on tasks and sessions restart independently.
- Metadata: Status/assignee/priority/outcome/checklist/notes/subtasks/handoff/blocking/type; no labels, estimates or custom fields.
- Durable states: Planned/Active/Done/Archived with transitions, restore semantics and daemon-persisted context across restarts.
- No labels/tags, estimates or epic views; single-product tracking only.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
Unattended triggers: Yes cross-checked
Unattended triggers
Typed value: Yes.
Daemon-owned automation runs interval, cron and one-time rules with notify plus one-time group-state/actor-control actions, completion semantics and Run-now; delivery retries are bounded and uncertain handoffs need explicit retry confirmation. No external event triggers exist. Automation needs the daemon host running (service manager for unattended); paused/stopped groups run nothing and missed fires are not replayed.
- Schedules: Interval (every N minutes), recurring (daily/weekly/monthly cron with timezone) and one-time (countdown/exact) rule triggers with max_runs.
- Events: IM messages, automation notices and Mail/reply notices move work; generic external event triggers (webhooks/repo events) do not exist.
- Recurring tasks: Cron recurring reminders/operations with timezone handling and progress/completion lists.
- Completion conditions: One-time rules complete and disable after firing; max_runs bounds repeats; past-due fires while paused/stopped do not run late.
- Scheduler location: The daemon owns scheduling; rule/automation state lives in groups/<id>/state/automation.json under CCCC_HOME.
- Overlaps: Resume restarts clocks without replaying missed fires; concurrent-fire overlap semantics are unstated.
- Retries: Bounded delivery retries with ambiguous settlement (no auto-retry of uncertain handoffs); Run now does not retry unknown outcomes; failed deliveries offer Retry.
- No external event triggers; unattended operation needs a supervisor.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Automation supports interval, cron and one-time triggers with optional max_runs; paused/stopped groups run none, idle groups suppress the built-in stand-up, resume restarts clocks.
Team roles: Yes cross-checked
Team roles
Typed value: Yes.
Per-group user/foreman/peer roles with a daemon-enforced permission matrix plus per-task assignees; account device membership and Connect member IDs; token-gated access with revocation (cut device, unlink, rotate). No organizations, no SSO/provisioning, no plan-scoped roles.
- Organization roles: Roles, organization_roles: bounded inspection found no organization entity or org-level roles; accounts link devices, not orgs.
- Project roles: Per-group user/foreman/peer roles with documented capabilities (foreman manages any actor, peers only self).
- Sso: Roles, sso: bounded search of docs, CLI and pinned sources found no single sign-on; sign-in is GitHub/Google/email-link plus local tokens.
- Provisioning: Roles, provisioning: no user provisioning/deprovisioning flow exists beyond device cut/unlink and token rotation.
- Unauthorized access: Loopback-only local principal, Admin-token gates for non-local, exhibit read-only; agent-runtime auth inherits providers.
- Plan scope: Roles, plan_scope: no plans or tiers exist, so no plan-scoped controls exist either.
- Group-local roles only; no org/SSO/provisioning.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
SSO: No cross-checked
SSO
Typed value: No.
Per-group user/foreman/peer roles with a daemon-enforced permission matrix plus per-task assignees; account device membership and Connect member IDs; token-gated access with revocation (cut device, unlink, rotate). No organizations, no SSO/provisioning, no plan-scoped roles.
- Organization roles: Roles, organization_roles: bounded inspection found no organization entity or org-level roles; accounts link devices, not orgs.
- Project roles: Per-group user/foreman/peer roles with documented capabilities (foreman manages any actor, peers only self).
- Sso: Roles, sso: bounded search of docs, CLI and pinned sources found no single sign-on; sign-in is GitHub/Google/email-link plus local tokens.
- Provisioning: Roles, provisioning: no user provisioning/deprovisioning flow exists beyond device cut/unlink and token rotation.
- Unauthorized access: Loopback-only local principal, Admin-token gates for non-local, exhibit read-only; agent-runtime auth inherits providers.
- Plan scope: Roles, plan_scope: no plans or tiers exist, so no plan-scoped controls exist either.
- Group-local roles only; no org/SSO/provisioning.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
Pull request workflow: No cross-checked
Pull request workflow
Typed value: No.
Git stays terminal- and agent-side: CCCC surfaces read-only diffs and conflict displays, documents branch/worktree conventions (including Claude background worktrees) and conflict inspection, but never stages, discards, commits, pushes, rebases or opens pull requests, and owns no PR state.
- Commits: Git workflow, commits: CCCC never commits; Changes is read-only and agents commit via their own tools.
- Pull requests: Git workflow, pull_requests: no product PR creation or management exists; agents discuss/ask for reviews in chat.
- Comments: Git workflow, comments: no PR review-comment integration exists; message reply/quote is the comment surface.
- Ci response: Git workflow, ci_response: no CI integration or response handling exists.
- Rebase: Git workflow, rebase: no rebase support exists.
- Conflicts: Conflict entries display with current-file inspection; resolution stays in terminal/agents.
- Product state: Read-only Git state view (Changes) with saved-content semantics; no product-owned PR/merge state.
- Read-only Git visibility; all mutation is runtime-side.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
Review interface: Yes cross-checked
Review interface
Typed value: Yes.
Diffs (read-only Changes view), documents (Files editor), screenshots/recordings attachments, HTML/Markdown previews, review tables and agent-reported test outputs persist linked to tasks and revisions; Presentation slots and pins form the cross-task gallery; completed child findings persist as outcomes and Agent State. No cross-task artifact search or prompt-leakage statement exists.
- Diffs: Changes view lists working-tree/staged/conflicts/untracked with unified/side-by-side diffs; read-only by design.
- Documents: Workspace editor with save/conflict/reload semantics, 1 MiB text limit, CSV/Markdown/Mermaid rendering, downloads.
- Screenshots: Image attachments with shared graphics viewer (fit/actual/zoom/pan), before/after review screenshots, Presentation images.
- Previews: Static isolated HTML previews, Markdown rendering, CSV tables, PDF/video native viewers, embedded browser views.
- Test outputs: Agent-reported validation evidence (e.g. pnpm test tallies) arrives as messages/outcomes; no product test-output artifact type.
- Reviewer outputs: Review tables, before/after evidence and outcomes persist in messages and tasks.
- Gallery: Presentation four-slot rail plus pins tab form the browsable cross-view collection linked to messages and tasks.
- Task links: Backlog topic links, handoff/blocking links, message/event links and Show-in-chat bind artifacts to tasks.
- Persistence: Ledger/blobs/context/tasks survive restarts; retained terminals/views expire by policy; copies carry history.
- Child findings: Peer/foreman findings return as replies, outcomes and Agent State reports, retained in the ledger.
- No gallery search; test evidence is message-grade; prompt exposure unstated.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
Product license: Apache-2.0: Yes. Apache-2.0 cross-checked
Product license: Apache-2.0
Typed value: Yes. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: MIT: No. Apache-2.0 cross-checked
Product license: MIT
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: GPL-3.0: No. Apache-2.0 cross-checked
Product license: GPL-3.0
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: GPL-3.0-or-later: No. Apache-2.0 cross-checked
Product license: GPL-3.0-or-later
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: AGPL-3.0: No. Apache-2.0 cross-checked
Product license: AGPL-3.0
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: AGPL-3.0-or-later: No. Apache-2.0 cross-checked
Product license: AGPL-3.0-or-later
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: Elastic License 2.0: No. Apache-2.0 cross-checked
Product license: Elastic License 2.0
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: Custom terms: No. Apache-2.0 cross-checked
Product license: Custom terms
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Product license: Proprietary: No. Apache-2.0 cross-checked
Product license: Proprietary
Typed value: No. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Self-hosting: Yes cross-checked
Self-hosting
Typed value: Yes.
Fully self-hosted on owned Linux/macOS/Windows or Docker with no required vendor services; optional Reach/account relay and user-chosen tunnels/VPNs only extend access. Copy Groups exports (zip) and imports durable state, CCCC_HOME backup/restore ports everything, formats are zip/JSONL/YAML/Markdown; account deletion cuts devices and removes the record while uninstalls intentionally keep data.
- Self hosting: Headless/systemd/container/Docker self-hosting with no vendor service required; local use needs no account.
- Export: Copy Groups export (ledger/history/agents/context/automation/settings) plus whole CCCC_HOME backup.
- Workspace files and live state do not port; copies exclude secrets.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
Free access: Yes cross-checked
Free access
Typed value: Yes.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: Free: Yes. Free cross-checked
Starting price: Free
Typed value: Yes. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: Up to $20/month: No. Free cross-checked
Starting price: Up to $20/month
Typed value: No. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: $21–40/month: No. Free cross-checked
Starting price: $21–40/month
Typed value: No. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: $41–100/month: No. Free cross-checked
Starting price: $41–100/month
Typed value: No. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: Over $100/month: No. Free cross-checked
Starting price: Over $100/month
Typed value: No. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Starting price: Contact for pricing: No. Free cross-checked
Starting price: Contact for pricing
Typed value: No. Free.
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Ongoing free: Free Apache-2.0 download/installers/pip/Docker/source with no paid tier in checked sources; local use needs no account.
- Trial: Free tier, trial: no trial exists; the product itself is free.
- Credits trial: Free tier, credits_trial: no credits trial exists.
- Self host cost: No license cost; the user provides hardware/VPS and provider subscriptions; Reach helper install is free.
- Paid only: Free tier, paid_only: not paid-only; everything runs free of charge.
- Early access: Free tier, early_access: no gated early access; RC channel is public and opt-in.
- Card required: Free tier, card_required: no card or billing exists anywhere in the checked sources.
- Usage caps: Free tier, usage_caps: no tier caps exist; incidental product limits (100 pins, bounded terminals/transcripts) are not billing caps.
- Paid evaluation: Free tier, paid_evaluation: no paid evaluation exists; evaluation is the free download.
- Free means the whole product; provider spend stays provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
Pricing model: Free, no paid plan: Yes cross-checked
Pricing model: Free, no paid plan
Typed value: Yes.
CCCC has no paid plan and no prices: free Apache-2.0 download (install script, pip wheel, Docker, source) with no currency, intervals, seats, overages, taxes or credits anywhere in the checked sources. Agent spend is the user own existing subscriptions or API accounts, handled provider-side.
- Credits: Price, credits: no credits are sold or mentioned; spend is provider-side.
- Model cost: Agent cost is the user own existing subscriptions or API accounts, handled provider-side; CCCC sells no tokens.
- Priced dimensions are absent because no paid plan exists; no purchase or billing test was run.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
Pricing model: Bring your own key: Yes cross-checked
Pricing model: Bring your own key
Typed value: Yes.
CCCC has no paid plan and no prices: free Apache-2.0 download (install script, pip wheel, Docker, source) with no currency, intervals, seats, overages, taxes or credits anywhere in the checked sources. Agent spend is the user own existing subscriptions or API accounts, handled provider-side.
- Credits: Price, credits: no credits are sold or mentioned; spend is provider-side.
- Model cost: Agent cost is the user own existing subscriptions or API accounts, handled provider-side; CCCC sells no tokens.
- Priced dimensions are absent because no paid plan exists; no purchase or billing test was run.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
Keeps going with laptop closed: Yes cross-checked
Keeps going with laptop closed
Typed value: Yes.
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
- Local: Daemon + agents run on the local machine with state in CCCC_HOME; localhost use needs no token.
- Byo remote: LAN binding, Cloudflare tunnel, Tailscale and Reach-published own host give remote/phone access to the same instance.
- Self hosted: Headless/systemd/container deployments documented with backup/restore runbook; Docker image ships agent CLIs.
- Managed cloud: Execution, managed_cloud: bounded inspection found no vendor cloud compute; Remote Access serves the user own machine and Managed Access is a tunnel, not hosted execution.
- Client disconnect: Closing the browser/IM never stops the daemon; reconnect resumes views, cursors and retained terminals from ledger authority.
- Other device continuation: Phone/browser/IM/Connect continue the same Groups from other devices; in-flight page drafts stay per browser.
- No vendor cloud; regions/resources/payer are user-side and unstated.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
Open-source license: Yes. Apache-2.0 cross-checked
Open-source license
Typed value: Yes. Apache-2.0.
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Client license: Web UI, CLI and daemon sources under root Apache-2.0 LICENSE; web bundle ships from the same licensed tree.
- Server license: Daemon/core/runtime crates under the same Apache-2.0 grant (workspace Apache-2.0, member crates inherit).
- Hosted account/relay service is closed with no parity statement.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
Any desktop app: Yes cross-checked
Any desktop app
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Cli: Product-owned cccc CLI covers daemon lifecycle, groups, actors, messaging, inbox, ledger search, IM, space, setup, update, web and mcp, plus an interactive terminal interface with status panel and Waiting-for-you rows.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
Any phone client: Yes cross-checked
Any phone client
Typed value: Yes.
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- Desktop os: Desktop OS coverage is the daemon/CLI plus browser: Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. Intel Macs stop at 0.4.37; Linux ARM64/Alpine/musl and native Windows ARM64 Python envs do not match current wheels.
- Mobile apps: No native iOS/Android app ships. The Web UI is responsive with phone layouts and a PWA manifest (standalone display); iPhone/iPad use is via browser or Home Screen web app. Phone Files view is read-only.
- Browser: The Web UI is the primary client at 127.0.0.1:8848/ui, usable over LAN/tunnel/Tailscale/Reach with an Admin token; read-only exhibit mode exists for sharing.
- Mobile browser: Mobile browsers work including swipe/pull-to-refresh/touch menus; Cloudflare Tunnel + Access recommended for phone access; iOS/iPadOS 16.4+ Home Screen apps get Web Push with 18.4+ link-opening.
- Editor extensions: Client access, editor extensions: bounded search of the site, docs, CLI reference and pinned repo found no editor/IDE extension; editing happens in the Web file browser or the agent runtime terminals.
- Task actions: Start, steer, approve, review and resume run from Web (run controls, in-place answers, diffs), the terminal panel (group/actor pages), the CLI (group/actor/send/reply/deliver) and IM (/send /status /pause /resume).
- Notifications: Web Push per browser (message addressed to user + open requests, bounded retries), IM bridge updates, automation notify rules with Mail/reply notices, terminal panel Waiting-for-you counts.
- Cross device continuity: The same Group is reachable from another device via LAN/tunnel/Tailscale/Reach remote access, IM bridges and Connect; drafts and reading places stay per browser (unsent drafts live only in the open page).
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
All public findings
Findings may summarize a source. They are not presented as verbatim statements.
cccc:ACC-01: Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect. cross-checked
cccc:ACC-01
Daemon + CLI run on Linux x86-64 (glibc 2.28+), Apple Silicon macOS 11+ and Windows x86-64; the React Web UI works in desktop and mobile browsers and as an installed standalone-display web app (PWA manifest, iOS Home Screen from 16.4 for push). No native desktop or mobile app and no editor extension ship. Start/steer/approve/review/resume run in Web, terminal panel, CLI and IM; Web Push plus IM/automation notices cover notifications; the same Group continues across devices via remote access or Connect.
- No native desktop/mobile apps or editor extension; phone Files read-only; drafts and reading places are per-browser, not synced.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: PWA manifest present: name CCCC, display standalone, icons; the Web UI is an installable standalone-display web app, not a native mobile/desktop client.
cccc:ACC-02: Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key. cross-checked
cccc:ACC-02
Each agent picks one of 21 named runtimes (Claude Code, Codex, Grok Build, OpenCode, Kilo, Copilot, Cursor, Devin, Kiro, Antigravity, Droid, Auggie, Hermes, Kimi, Cline, Amp, Pi, DeepSeek, ChatGPT Web, ChatGPT Dots, Grok Bot) or any command via custom, and mixed runtimes share one Group. Model choice stays runtime-side (model-label is declarative only). CCCC injects its MCP into managed sessions and hooks native ones; provider logins stay with the runtimes and CCCC needs no provider API key.
- Model choice and provider login stay with each runtime; actor model-label is declarative and unchecked.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Browser-delivered ChatGPT agent with remote MCP connector: sign-in + shared connector + bound conversation; bounded repository inspection and scoped editing; unverified delivery/unsent-draft boundaries listed.
-
Finding: Agents invoke skill: capabilities via slash dispatch with fixed send mode; skills are installed capability packs, not free-form plugins.
-
Finding: Capabilities install from explicit local path, URL or GitHub target with recorded provenance; remote search results stay unverified until imported from a trusted origin.
cccc:ACC-03: Agents run on the user own existing subscriptions and logins: connected sessions keep their app process, account, model, approvals and terminal, and CCCC needs no provider API key. Managed agents are launched from the user-installed CLIs. No vendor credits, paid CCCC plans or card exist; per-actor private environment supports different model/API stacks per agent. The only account is the optional free CCCC account for Connect/Reach; local use needs none. cross-checked
cccc:ACC-03
Agents run on the user own existing subscriptions and logins: connected sessions keep their app process, account, model, approvals and terminal, and CCCC needs no provider API key. Managed agents are launched from the user-installed CLIs. No vendor credits, paid CCCC plans or card exist; per-actor private environment supports different model/API stacks per agent. The only account is the optional free CCCC account for Connect/Reach; local use needs none.
- No CCCC-side billing or credits; all inference spend stays on the user runtime accounts.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Browser-delivered ChatGPT agent with remote MCP connector: sign-in + shared connector + bound conversation; bounded repository inspection and scoped editing; unverified delivery/unsent-draft boundaries listed.
-
Finding: Chet global assistant (text + experimental Codex/Grok voice using their logins, no extra API key) and Voice Secretary dictation/documents/live voice with local or cloud ASR options.
-
Finding: Group Space binds NotebookLM work/memory notebooks via Google credential; ingestion/query/jobs are explicit agent/CLI operations; automatic two-way sync retired; disconnect keeps core workflows running.
cccc:ACC-04: Install (shell/PowerShell/pip/Docker/source), run cccc, create a Group, add an installed signed-in agent and send the first message. Any local directory attaches as a scope with no Git requirement; Git clones get read-only diffs (CCCC never stages/commits). Prerequisites are an agent CLI install + sign-in (plus Node 24 for JS code mode, Xvfb for projected browsers). Managed cloud does not exist; a BYO VPS path runs headless under a supervisor with LAN/tunnel/Tailscale/Reach access (Reach helper Linux/macOS only). cross-checked
cccc:ACC-04
Install (shell/PowerShell/pip/Docker/source), run cccc, create a Group, add an installed signed-in agent and send the first message. Any local directory attaches as a scope with no Git requirement; Git clones get read-only diffs (CCCC never stages/commits). Prerequisites are an agent CLI install + sign-in (plus Node 24 for JS code mode, Xvfb for projected browsers). Managed cloud does not exist; a BYO VPS path runs headless under a supervisor with LAN/tunnel/Tailscale/Reach access (Reach helper Linux/macOS only).
- No managed-cloud option; Reach remote publishing is a Linux/macOS preview.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Documented path from fresh install to first task: install, cccc launch, create Group, add agent, send message.
-
Finding: Web flow: open local workspace in browser, create Group from sidebar, add installed agent, start it, send/communicate.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
cccc:ACC-05: A complete product-owned CLI plus MCP surfaces (daemon tools over stdio for agents; Python/TypeScript/Rust SDK clients for integrators; documented Web API endpoints), but no generic public REST API with stability promises and no generic outgoing webhooks. Create/inspect/steer/stop/review run through CLI, MCP, Web and IM; auth is loopback-local principal plus Admin Access Tokens for non-local use. cross-checked
cccc:ACC-05
A complete product-owned CLI plus MCP surfaces (daemon tools over stdio for agents; Python/TypeScript/Rust SDK clients for integrators; documented Web API endpoints), but no generic public REST API with stability promises and no generic outgoing webhooks. Create/inspect/steer/stop/review run through CLI, MCP, Web and IM; auth is loopback-local principal plus Admin Access Tokens for non-local use.
- No versioned public REST API or generic webhooks; automation-capable surface is CLI/MCP/Web/IM.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
cccc:ACC-06: Capabilities install as mounted skill packs from explicit local paths, URLs or GitHub targets with provenance; slash skills dispatch them. CCCC sets up its own MCP into most runtimes itself (prompt-assisted where the runtime requires it) and governs tool visibility with a foreman/user allowlist. No external-MCP consumption by CCCC itself and no plugin framework beyond capability packs were established; repo instructions import through the AGENTS.md editor, watched instruction files and cccc_file. cross-checked
cccc:ACC-06
Capabilities install as mounted skill packs from explicit local paths, URLs or GitHub targets with provenance; slash skills dispatch them. CCCC sets up its own MCP into most runtimes itself (prompt-assisted where the runtime requires it) and governs tool visibility with a foreman/user allowlist. No external-MCP consumption by CCCC itself and no plugin framework beyond capability packs were established; repo instructions import through the AGENTS.md editor, watched instruction files and cccc_file.
- Capability governance is tool visibility, not process sandboxing; no third-party MCP consumption.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: AGENTS.md/CLAUDE.md watched files with change nudges, per-group protocol reference customization, Builder + Reviewer pair pattern, code review agents recommended.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
-
Finding: Browser-delivered ChatGPT agent with remote MCP connector: sign-in + shared connector + bound conversation; bounded repository inspection and scoped editing; unverified delivery/unsent-draft boundaries listed.
-
Finding: Group Space binds NotebookLM work/memory notebooks via Google credential; ingestion/query/jobs are explicit agent/CLI operations; automatic two-way sync retired; disconnect keeps core workflows running.
-
Finding: Agents invoke skill: capabilities via slash dispatch with fixed send mode; skills are installed capability packs, not free-form plugins.
-
Finding: Capabilities install from explicit local path, URL or GitHub target with recorded provenance; remote search results stay unverified until imported from a trusted origin.
cccc:ACC-07: No installable desktop or native mobile client ships: access is the React+TypeScript+Vite+Tailwind+xterm Web UI in desktop/mobile browsers (installable as a standalone-display PWA) plus the Rust CLI/daemon and terminal panel. Implementation evidence is pinned source: web/package.json dependency set, PWA manifest, Rust workspace members. No hybrid wrappers were found. cross-checked
cccc:ACC-07
No installable desktop or native mobile client ships: access is the React+TypeScript+Vite+Tailwind+xterm Web UI in desktop/mobile browsers (installable as a standalone-display PWA) plus the Rust CLI/daemon and terminal panel. Implementation evidence is pinned source: web/package.json dependency set, PWA manifest, Rust workspace members. No hybrid wrappers were found.
- Browser+PWA+CLI only; no native or hybrid clients.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
-
Finding: PWA manifest present: name CCCC, display standalone, icons; the Web UI is an installable standalone-display web app, not a native mobile/desktop client.
cccc:PLAN-01: Groups, tasks and backlog entries are created natively by humans (composer, CLI, Web) with durable task/event IDs, and the Group ledger is the authoritative record for product-owned work. There is no import from Linear/GitHub/Jira or other external trackers and no chat/Git triggers; CCCC-to-CCCC Connect relays messages between instances rather than syncing trackers, so no external round-trip exists. cross-checked
cccc:PLAN-01
Groups, tasks and backlog entries are created natively by humans (composer, CLI, Web) with durable task/event IDs, and the Group ledger is the authoritative record for product-owned work. There is no import from Linear/GitHub/Jira or other external trackers and no chat/Git triggers; CCCC-to-CCCC Connect relays messages between instances rather than syncing trackers, so no external round-trip exists.
- Self-contained issue tracking; no external tracker integration of any kind.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Web flow: open local workspace in browser, create Group from sidebar, add installed agent, start it, send/communicate.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:PLAN-02: Tasks carry title, status, assignee, priority, outcome, checklist, notes, subtasks and handoff/blocking links, editable in the Web board, CLI-tracked sends and agent tools with optimistic-concurrency revisions. Plans amend freely mid-run, link to backlog topics and handoffs, and persist across restarts in the daemon-owned context store. cross-checked
cccc:PLAN-02
Tasks carry title, status, assignee, priority, outcome, checklist, notes, subtasks and handoff/blocking links, editable in the Web board, CLI-tracked sends and agent tools with optimistic-concurrency revisions. Plans amend freely mid-run, link to backlog topics and handoffs, and persist across restarts in the daemon-owned context store.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:PLAN-03: Tasks form a parent/subtask tree with assignees and blocked_by/waiting-on attention states derived into blocked badges (recomputed, cf. fix #201), linked to backlog topics that outlive them. No epic entity, readiness gate or dependency enforcement exists: type requirements never block a change and Mail never starts an agent. No rigid workflow is required by design. cross-checked
cccc:PLAN-03
Tasks form a parent/subtask tree with assignees and blocked_by/waiting-on attention states derived into blocked badges (recomputed, cf. fix #201), linked to backlog topics that outlive them. No epic entity, readiness gate or dependency enforcement exists: type requirements never block a change and Mail never starts an agent. No rigid workflow is required by design.
- Derived (not enforced) blockers; no epics, gates or mandatory workflows.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: CCCC is the coordination layer, not a replacement for agents; complements native agent teams, parallel task runners and IM gateways rather than replacing them.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:PLAN-04: Repository instructions live in the workspace AGENTS.md editor plus watched AGENTS/CLAUDE variants with change nudges; project docs in Files; working memory in Tasks/context/Agent State and the NotebookLM work/memory notebooks; skills as installable capability packs with self-evolving candidates surfaced per Group. Provenance is ledger attribution with staleness marks; correction/deletion run through editors, archive/delete and version conflicts; no spec-drift detection exists. cross-checked
cccc:PLAN-04
Repository instructions live in the workspace AGENTS.md editor plus watched AGENTS/CLAUDE variants with change nudges; project docs in Files; working memory in Tasks/context/Agent State and the NotebookLM work/memory notebooks; skills as installable capability packs with self-evolving candidates surfaced per Group. Provenance is ledger attribution with staleness marks; correction/deletion run through editors, archive/delete and version conflicts; no spec-drift detection exists.
- Chet cross-group context is instruction-level, not isolated; memory search/update tooling retired.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: AGENTS.md/CLAUDE.md watched files with change nudges, per-group protocol reference customization, Builder + Reviewer pair pattern, code review agents recommended.
-
Finding: Group Space binds NotebookLM work/memory notebooks via Google credential; ingestion/query/jobs are explicit agent/CLI operations; automatic two-way sync retired; disconnect keeps core workflows running.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Agents invoke skill: capabilities via slash dispatch with fixed send mode; skills are installed capability packs, not free-form plugins.
-
Finding: Capabilities install from explicit local path, URL or GitHub target with recorded provenance; remote search results stay unverified until imported from a trusted origin.
cccc:PLAN-05: Agents read the workspace through Files, message attachments (images/video/PDF/audio), Mermaid/PDF viewers, repo inspection tools and explicit NotebookLM ingestion (files, pasted text, URLs, YouTube, Drive). Bounded repo inspection and scoped editing exist for web runtimes. No product-wide code indexing, citation model, conflict detection or review-time enforcement was established. cross-checked
cccc:PLAN-05
Agents read the workspace through Files, message attachments (images/video/PDF/audio), Mermaid/PDF viewers, repo inspection tools and explicit NotebookLM ingestion (files, pasted text, URLs, YouTube, Drive). Bounded repo inspection and scoped editing exist for web runtimes. No product-wide code indexing, citation model, conflict detection or review-time enforcement was established.
- Retrieval is tool- and file-scoped; no index, citations model or enforcement.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Browser-delivered ChatGPT agent with remote MCP connector: sign-in + shared connector + bound conversation; bounded repository inspection and scoped editing; unverified delivery/unsent-draft boundaries listed.
-
Finding: Chet global assistant (text + experimental Codex/Grok voice using their logins, no extra API key) and Voice Secretary dictation/documents/live voice with local or cloud ASR options.
-
Finding: Group Space binds NotebookLM work/memory notebooks via Google credential; ingestion/query/jobs are explicit agent/CLI operations; automatic two-way sync retired; disconnect keeps core workflows running.
-
Pinned-tree negative search for repository indexing
Finding: Bounded inspection of the pinned tree (crates/*/src, web/src, docs/**) finds no repository-indexing service or index-backed retrieval implementation; retrieval is tool- and attachment-scoped.
cccc:PLAN-06: A complete native task system: create/edit/discuss/prioritize/assign/search/filter/archive/reopen in the board, idea parking in the project BACKLOG.md, explicit dispatch via tracked-send, durable Planned/Active/Done states and task/session separation with per-task metadata. No labels/tags, no epic views, no external tracker: nothing external is needed or integrated. cross-checked
cccc:PLAN-06
A complete native task system: create/edit/discuss/prioritize/assign/search/filter/archive/reopen in the board, idea parking in the project BACKLOG.md, explicit dispatch via tracked-send, durable Planned/Active/Done states and task/session separation with per-task metadata. No labels/tags, no epic views, no external tracker: nothing external is needed or integrated.
- No labels/tags, estimates or epic views; single-product tracking only.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
cccc:PLAN-07: Agents share the same durable task IDs humans use: list/get via cccc_task-class tools, create via tracked delegation, update with optimistic-concurrency batches under foreman/peer role gates, assign/state/subtask/result through context ops, all visible in the Web board with ledger attribution. Self-claim and agent-set dependencies are only partly established; conflicts surface as explicit task_conflict errors. cross-checked
cccc:PLAN-07
Agents share the same durable task IDs humans use: list/get via cccc_task-class tools, create via tracked delegation, update with optimistic-concurrency batches under foreman/peer role gates, assign/state/subtask/result through context ops, all visible in the Web board with ledger attribution. Self-claim and agent-set dependencies are only partly established; conflicts surface as explicit task_conflict errors.
- Agent self-claim and agent-set dependencies only partly established.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:PLAN-08: Long-lived requirements persist in the project BACKLOG.md (topics outlive tasks with outcome write-back) and in AGENTS.md guidance; both are human-editable with version/hash guards. Agents write outcomes and notes and read topics/files via tools, with later retrieval through ledger search and context. No formal requirement objects, approval records or drift detection exist. cross-checked
cccc:PLAN-08
Long-lived requirements persist in the project BACKLOG.md (topics outlive tasks with outcome write-back) and in AGENTS.md guidance; both are human-editable with version/hash guards. Agents write outcomes and notes and read topics/files via tools, with later retrieval through ledger search and context. No formal requirement objects, approval records or drift detection exist.
- File-based persistence, not a requirements database; no drift detection or approval records.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: AGENTS.md/CLAUDE.md watched files with change nudges, per-group protocol reference customization, Builder + Reviewer pair pattern, code review agents recommended.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:COORD-01: Several agents run concurrently in a Group and several Groups run side by side; the sidebar, terminal panel and status counts show running, waiting, failed and finished work across them. No concurrency quotas are documented; observed-concurrency evidence is limited to status/running counts rather than measured parallelism. cross-checked
cccc:COORD-01
Several agents run concurrently in a Group and several Groups run side by side; the sidebar, terminal panel and status counts show running, waiting, failed and finished work across them. No concurrency quotas are documented; observed-concurrency evidence is limited to status/running counts rather than measured parallelism.
- No quotas; concurrency visibility is status-level, not measured.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
cccc:COORD-02: Users and the foreman assign and route work (@mentions, default_send_to policy, tracked-send ownership); peers manage only themselves under daemon RBAC. Dispatch is guarded by idempotency keys and single-writer rules. No queues and no duplicate-prevention beyond idempotency were established. cross-checked
cccc:COORD-02
Users and the foreman assign and route work (@mentions, default_send_to policy, tracked-send ownership); peers manage only themselves under daemon RBAC. Dispatch is guarded by idempotency keys and single-writer rules. No queues and no duplicate-prevention beyond idempotency were established.
- No queues; claiming semantics thin.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:COORD-03: The foreman coordinates peers in a daemon-owned Group: handoffs via direct agent-to-agent messages with delivery/read/reply facts, per-agent runtime/model-label/private-env configuration, mixed-harness teams, live activity plus retained terminals, and retained results in ledger/tasks/Agent State. Follow-up runs through replies, retries, new sessions and restarts across CLI, MCP, Web and IM. cross-checked
cccc:COORD-03
The foreman coordinates peers in a daemon-owned Group: handoffs via direct agent-to-agent messages with delivery/read/reply facts, per-agent runtime/model-label/private-env configuration, mixed-harness teams, live activity plus retained terminals, and retained results in ledger/tasks/Agent State. Follow-up runs through replies, retries, new sessions and restarts across CLI, MCP, Web and IM.
- Effort/turn budgets are runtime-side; auth partly inherits runtimes.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
cccc:COORD-04: Daemon-owned automation runs interval, cron and one-time rules with notify plus one-time group-state/actor-control actions, completion semantics and Run-now; delivery retries are bounded and uncertain handoffs need explicit retry confirmation. No external event triggers exist. Automation needs the daemon host running (service manager for unattended); paused/stopped groups run nothing and missed fires are not replayed. cross-checked
cccc:COORD-04
Daemon-owned automation runs interval, cron and one-time rules with notify plus one-time group-state/actor-control actions, completion semantics and Run-now; delivery retries are bounded and uncertain handoffs need explicit retry confirmation. No external event triggers exist. Automation needs the daemon host running (service manager for unattended); paused/stopped groups run nothing and missed fires are not replayed.
- No external event triggers; unattended operation needs a supervisor.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
-
Finding: Automation supports interval, cron and one-time triggers with optional max_runs; paused/stopped groups run none, idle groups suppress the built-in stand-up, resume restarts clocks.
cccc:COORD-05: Pause/stop/resume/steer from Web run controls, CLI and IM; Web Push, IM and automation notices plus in-place pending decisions under Waiting for you; delivery failures surface with reasons and Retry; chat and terminal views share the ledger with documented native-session limits; history persists in the ledger, tasks and pins. cross-checked
cccc:COORD-05
Pause/stop/resume/steer from Web run controls, CLI and IM; Web Push, IM and automation notices plus in-place pending decisions under Waiting for you; delivery failures surface with reasons and Retry; chat and terminal views share the ledger with documented native-session limits; history persists in the ledger, tasks and pins.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:COORD-06: The sidebar lists every Group with lifecycle status and Waiting-for-you counts (including archived totals), each Group page shows run controls, task-state counts and attention filters, and terminals page per agent with waiting/stuck indicators. Subtasks and blocked_by show on cards, but there is no fleet-level dependency, workload or navigation-error view and no fixture/demo mode. cross-checked
cccc:COORD-06
The sidebar lists every Group with lifecycle status and Waiting-for-you counts (including archived totals), each Group page shows run controls, task-state counts and attention filters, and terminals page per agent with waiting/stuck indicators. Subtasks and blocked_by show on cards, but there is no fleet-level dependency, workload or navigation-error view and no fixture/demo mode.
- No fleet dependency/workload views; navigation-error behavior unstated.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
cccc:COORD-07: A plain send starts work with no mandatory workflow or preset; snippets and the opt-in stand-up are optional. Agents reconfigure per runtime/command/private-env mid-run, roles show live (foreman badge, status, activity), and ledger/task/pin links retain relationships. Management is discoverable via docs, /help, doctor and IM commands; local use is passwordless while remote actions need tokens. cross-checked
cccc:COORD-07
A plain send starts work with no mandatory workflow or preset; snippets and the opt-in stand-up are optional. Agents reconfigure per runtime/command/private-env mid-run, roles show live (foreman badge, status, activity), and ledger/task/pin links retain relationships. Management is discoverable via docs, /help, doctor and IM commands; local use is passwordless while remote actions need tokens.
- Optional snippets only; no workflow presets or NL workflow definitions.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Web flow: open local workspace in browser, create Group from sidebar, add installed agent, start it, send/communicate.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
cccc:RUN-01: Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started. cross-checked
cccc:RUN-01
Execution is local on the user machine (plus self-hosted headless/Docker/service-manager deployments and BYO remote access to that machine over LAN, tunnel, Tailscale or Reach). There is no vendor cloud compute: Reach publishes the user own host (uploads neither repo nor ledger) and Connect relays only messages/files. Work survives client disconnects and continues from other devices; disconnecting the launching terminal stops what it started.
- No vendor cloud; regions/resources/payer are user-side and unstated.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:RUN-02: Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies. cross-checked
cccc:RUN-02
Agents share the Group checkout by default with workspace-scoped file checks; Claude managed sessions additionally default to background worktrees unless configured otherwise. No product containers, VMs, OS sandboxes or process/network isolation exist — capability allowlists govern tool visibility, not agent processes — and the Web UI is explicitly high-privilege. Secrets stay in CCCC_HOME with owner-only permissions and out of copies.
- Shared-checkout default; isolation is scoping, not sandboxing.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
-
Finding: Group Space binds NotebookLM work/memory notebooks via Google credential; ingestion/query/jobs are explicit agent/CLI operations; automatic two-way sync retired; disconnect keeps core workflows running.
cccc:RUN-03: The append-only ledger plus context/mail cursors and delivery facts is the recovery authority: daemon restarts settle claims, resume automation clocks and restore views/terminals; actor/group/daemon restart ladders and native-session recovery are documented. Transient terminal modes are not transferred and uncertain handoffs are never auto-replayed; idempotency prevents double-apply and the operations runbook covers manual repair. cross-checked
cccc:RUN-03
The append-only ledger plus context/mail cursors and delivery facts is the recovery authority: daemon restarts settle claims, resume automation clocks and restore views/terminals; actor/group/daemon restart ladders and native-session recovery are documented. Transient terminal modes are not transferred and uncertain handoffs are never auto-replayed; idempotency prevents double-apply and the operations runbook covers manual repair.
- Transient UI/provider state is not reconstructed; uncertainty never auto-replays.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:RUN-04: Agent CLIs stay user-installed with doctor/runtime-list diagnostics and per-runtime setup repair; ports are configurable with collision guidance; actor secrets and remote-access setup are documented. No dependency provisioning, dev services, snapshots or automatic repair of environments exists, and worktree/duplicate hygiene stays manual. cross-checked
cccc:RUN-04
Agent CLIs stay user-installed with doctor/runtime-list diagnostics and per-runtime setup repair; ports are configurable with collision guidance; actor secrets and remote-access setup are documented. No dependency provisioning, dev services, snapshots or automatic repair of environments exists, and worktree/duplicate hygiene stays manual.
- User owns dependencies and hygiene; product diagnoses more than it repairs.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:RUN-05: No time/turn/token/spend limits or provider-failure policies gate agent work; costs stay provider-side and untracked. Bounded delivery retries, max_runs automation bounds, transport timeouts and failure surfacing with Retry exist, but nothing enforces a stop on spend or sized budgets. cross-checked
cccc:RUN-05
No time/turn/token/spend limits or provider-failure policies gate agent work; costs stay provider-side and untracked. Bounded delivery retries, max_runs automation bounds, transport timeouts and failure surfacing with Retry exist, but nothing enforces a stop on spend or sized budgets.
- Advisory and bounded-retry only; no enforced budgets of any kind.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:RUN-06: Daemon/terminal/ledger logs with tail/search/export, delivery failure reasons with diagnostics, task/attention statistics on the board and inbox summaries, lifecycle state badges and Waiting-for-you approval counts all run from ledger truth with reconnect-safe cursors. No distributed tracing, cost/time aggregates or measured performance figures exist. cross-checked
cccc:RUN-06
Daemon/terminal/ledger logs with tail/search/export, delivery failure reasons with diagnostics, task/attention statistics on the board and inbox summaries, lifecycle state badges and Waiting-for-you approval counts all run from ledger truth with reconnect-safe cursors. No distributed tracing, cost/time aggregates or measured performance figures exist.
- Status truth is strong; analytics and measured performance are absent.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
-
Finding: Support via GitHub issues with version/OS/command/repro template, cccc doctor diagnostics, Discord and Telegram community channels.
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
cccc:RUN-07: CCCC tracks no model, API or compute spend anywhere: no per-task/session/child/epic costs, no harness aggregation, no billed-vs-estimated reconciliation, no token accounting, no budgets/alerts/exports and no drill-down. Cost visibility is limited to incidental provider-side figures inside runtime terminals; inference spend stays entirely on the user own runtime accounts. cross-checked
cccc:RUN-07
CCCC tracks no model, API or compute spend anywhere: no per-task/session/child/epic costs, no harness aggregation, no billed-vs-estimated reconciliation, no token accounting, no budgets/alerts/exports and no drill-down. Cost visibility is limited to incidental provider-side figures inside runtime terminals; inference spend stays entirely on the user own runtime accounts.
- No cost dimension of any kind; provider-side spend is invisible to CCCC.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
cccc:RUN-08: No compute-aware scheduling exists: no concurrency limits, queues, CPU/RAM awareness, backpressure, check reuse/dedup, bounded verification or fairness policies. Cancellation is the one supported control (cancel-reply, stop/restart, pause delivery); recovery reuses the ledger authority rather than scheduled capacity. cross-checked
cccc:RUN-08
No compute-aware scheduling exists: no concurrency limits, queues, CPU/RAM awareness, backpressure, check reuse/dedup, bounded verification or fairness policies. Cancellation is the one supported control (cancel-reply, stop/restart, pause delivery); recovery reuses the ledger authority rather than scheduled capacity.
- Coordination without compute scheduling; cancellation only.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Deliveries carry bounded retry limits; daemon-settled claims without outcome become ambiguous and are not auto-retried; failed provider turns release the next delivery after their retry limit.
cccc:TEAM-01: Group state is shared to every holder of access (remote browsers, IM chats, read-only exhibits) with message discussion, reply/quote/forward comments and ledger attribution, but live human presence and simultaneous multi-human co-prompting do not exist: Connect links instances for message exchange rather than shared steering, and ordering/conflict handling for co-prompts is absent. cross-checked
cccc:TEAM-01
Group state is shared to every holder of access (remote browsers, IM chats, read-only exhibits) with message discussion, reply/quote/forward comments and ledger attribution, but live human presence and simultaneous multi-human co-prompting do not exist: Connect links instances for message exchange rather than shared steering, and ordering/conflict handling for co-prompts is absent.
- Single-user-oriented sharing; no live co-steering.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
cccc:TEAM-02: Per-group user/foreman/peer roles with a daemon-enforced permission matrix plus per-task assignees; account device membership and Connect member IDs; token-gated access with revocation (cut device, unlink, rotate). No organizations, no SSO/provisioning, no plan-scoped roles. cross-checked
cccc:TEAM-02
Per-group user/foreman/peer roles with a daemon-enforced permission matrix plus per-task assignees; account device membership and Connect member IDs; token-gated access with revocation (cut device, unlink, rotate). No organizations, no SSO/provisioning, no plan-scoped roles.
- Group-local roles only; no org/SSO/provisioning.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:TEAM-03: User decisions gate progress through in-place question answers (recorded, requester-verified, scoped to the open request) and tracked outcomes; capability allowlists govern tool visibility while runtime tool approvals stay inherited or deliberately bypassed with warnings. No network/secret approval flows exist; temporary access (tunnels, shares) is cut explicitly; exhibit mode and conservative safety are the safer defaults. cross-checked
cccc:TEAM-03
User decisions gate progress through in-place question answers (recorded, requester-verified, scoped to the open request) and tracked outcomes; capability allowlists govern tool visibility while runtime tool approvals stay inherited or deliberately bypassed with warnings. No network/secret approval flows exist; temporary access (tunnels, shares) is cut explicitly; exhibit mode and conservative safety are the safer defaults.
- Decision gating is message-level; process-level approvals stay runtime-side.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: MCP capability governance: packaged default + user overlay, indexed/mounted/enabled/pinned/blocked levels, foreman-or-user group scope vs user-only global scope, external capability safety mode; explicitly not a sandbox for agent processes.
cccc:TEAM-04: Every action, approval, edit, discussion and decision lands attributed in the append-only ledger, persists across users/sessions/restarts, and exports through Copy Groups and ledger search. The ledger is append-only with verification/compaction (not a signed audit log); pins/tasks are mutable through guarded edits. cross-checked
cccc:TEAM-04
Every action, approval, edit, discussion and decision lands attributed in the append-only ledger, persists across users/sessions/restarts, and exports through Copy Groups and ledger search. The ledger is append-only with verification/compaction (not a signed audit log); pins/tasks are mutable through guarded edits.
- Durable and attributed, but not a signed audit trail.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:TEAM-05: Reviewers work from Waiting for you with counts, read review tables and before/after evidence in chat, navigate diffs in Changes, reply/quote/forward with context, and delegate fixes via tracked tasks; drafts, filters and reading places survive context switches and phone layouts adapt. No formal request-changes state, no exact-revision approvals, no keyboard-shortcut system and no effort estimates exist. cross-checked
cccc:TEAM-05
Reviewers work from Waiting for you with counts, read review tables and before/after evidence in chat, navigate diffs in Changes, reply/quote/forward with context, and delegate fixes via tracked tasks; drafts, filters and reading places survive context switches and phone layouts adapt. No formal request-changes state, no exact-revision approvals, no keyboard-shortcut system and no effort estimates exist.
- Strong review reading; informal review actions.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
cccc:TEAM-06: A second authorized person continues the same work from another device over remote access (same Group, history, decisions, artifacts, terminal context) or via Connect/IM from their own instance (message-level continuity). Owner disconnect never blocks: the daemon persists, answers record, and native sessions keep their apps. No per-user ownership transfer or harness-transfer ceremony exists beyond switching runtimes. cross-checked
cccc:TEAM-06
A second authorized person continues the same work from another device over remote access (same Group, history, decisions, artifacts, terminal context) or via Connect/IM from their own instance (message-level continuity). Owner disconnect never blocks: the daemon persists, answers record, and native sessions keep their apps. No per-user ownership transfer or harness-transfer ceremony exists beyond switching runtimes.
- Takeover is shared-access continuity, not a handover flow.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:SHIP-01: Diffs (read-only Changes view), documents (Files editor), screenshots/recordings attachments, HTML/Markdown previews, review tables and agent-reported test outputs persist linked to tasks and revisions; Presentation slots and pins form the cross-task gallery; completed child findings persist as outcomes and Agent State. No cross-task artifact search or prompt-leakage statement exists. cross-checked
cccc:SHIP-01
Diffs (read-only Changes view), documents (Files editor), screenshots/recordings attachments, HTML/Markdown previews, review tables and agent-reported test outputs persist linked to tasks and revisions; Presentation slots and pins form the cross-task gallery; completed child findings persist as outcomes and Agent State. No cross-task artifact search or prompt-leakage statement exists.
- No gallery search; test evidence is message-grade; prompt exposure unstated.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
-
Finding: Chet global assistant (text + experimental Codex/Grok voice using their logins, no extra API key) and Voice Secretary dictation/documents/live voice with local or cloud ASR options.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:SHIP-02: CCCC enforces no verification or spec gates: no test/lint/type/CI/review/spec checks, no required checks, no failure blocking and no configured or enforced policy. Tracked outcomes record validation evidence as text and reviewers iterate in chat, but a done task without an outcome only gets a suggestion and type requirements never block. cross-checked
cccc:SHIP-02
CCCC enforces no verification or spec gates: no test/lint/type/CI/review/spec checks, no required checks, no failure blocking and no configured or enforced policy. Tracked outcomes record validation evidence as text and reviewers iterate in chat, but a done task without an outcome only gets a suggestion and type requirements never block.
- Evidence recorded, never gated.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
cccc:SHIP-03: Git stays terminal- and agent-side: CCCC surfaces read-only diffs and conflict displays, documents branch/worktree conventions (including Claude background worktrees) and conflict inspection, but never stages, discards, commits, pushes, rebases or opens pull requests, and owns no PR state. cross-checked
cccc:SHIP-03
Git stays terminal- and agent-side: CCCC surfaces read-only diffs and conflict displays, documents branch/worktree conventions (including Claude background worktrees) and conflict inspection, but never stages, discards, commits, pushes, rebases or opens pull requests, and owns no PR state.
- Read-only Git visibility; all mutation is runtime-side.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
cccc:SHIP-04: CCCC owns no merge or deployment authority: no approve/merge/release/deploy actions, no exact-revision checks and no authority enforcement exist. Delivery authority stays with the user in their terminal, provider and hosting; the product coordinates and records instead. cross-checked
cccc:SHIP-04
CCCC owns no merge or deployment authority: no approve/merge/release/deploy actions, no exact-revision checks and no authority enforcement exist. Delivery authority stays with the user in their terminal, provider and hosting; the product coordinates and records instead.
- Coordination only; authority stays outside the product.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
cccc:SHIP-05: Delivery status is fully visible (runtime.delivery facts, lifecycle states, Waiting for you); outcomes, replies and attachments persist as retained evidence. But no commit/PR/build/deployment links, rollback/revert or stale-approval invalidation exist — recovery is backup restore or group-copy re-import, and concurrent saves conflict rather than invalidate. cross-checked
cccc:SHIP-05
Delivery status is fully visible (runtime.delivery facts, lifecycle states, Waiting for you); outcomes, replies and attachments persist as retained evidence. But no commit/PR/build/deployment links, rollback/revert or stale-approval invalidation exist — recovery is backup restore or group-copy re-import, and concurrent saves conflict rather than invalidate.
- Status and evidence without delivery mechanics.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:SHIP-06: High-volume shipping runs through Waiting for you, the board with attention filters, summaries in Agent State/stand-ups, review findings as tables, Presentation/pin artifacts and next-action surfacing — but with no unified review queue, risk/size filters, exact-revision approvals, dependency merge queues or check dedup. Request-changes and stale handling stay informal. cross-checked
cccc:SHIP-06
High-volume shipping runs through Waiting for you, the board with attention filters, summaries in Agent State/stand-ups, review findings as tables, Presentation/pin artifacts and next-action surfacing — but with no unified review queue, risk/size filters, exact-revision approvals, dependency merge queues or check dedup. Request-changes and stale handling stay informal.
- Review coordination without release mechanics.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Daemon task store supports planned/active/done/archived statuses, blocked/waiting_user/handoff/unassigned attention filters, assignee/query filtering and optimistic tasks_version concurrency.
-
Finding: Task mutations run as validated batches with compare-and-swap revisions and explicit task_conflict errors; role-gated so agent task operations follow the foreman/peer permission policy.
cccc:OWN-01: Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture. cross-checked
cccc:OWN-01
Apache-2.0 across the repository: root LICENSE grant covers daemon, Web UI, CLI, runtimes and docs; Cargo workspace and Python package metadata declare Apache-2.0. Source is fully visible at the pinned v0.4.55 commit; the hosted account/Connect service and the separate SDK repo are outside the licensed-repo parity picture.
- Hosted account/relay service is closed with no parity statement.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Canonical repo ChesterRa/cccc: 1309 stars observed 2026-11-13, Apache-2.0, homepage https://cccc.sh/, Rust, topics include claude-code, codex-cli, multi-agent, local-first.
-
Finding: Latest stable release v0.4.55 published 2026-11-12, with near-daily prior releases (v0.4.50–v0.4.54 within the preceding week).
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
cccc:OWN-02: Fully self-hosted on owned Linux/macOS/Windows or Docker with no required vendor services; optional Reach/account relay and user-chosen tunnels/VPNs only extend access. Copy Groups exports (zip) and imports durable state, CCCC_HOME backup/restore ports everything, formats are zip/JSONL/YAML/Markdown; account deletion cuts devices and removes the record while uninstalls intentionally keep data. cross-checked
cccc:OWN-02
Fully self-hosted on owned Linux/macOS/Windows or Docker with no required vendor services; optional Reach/account relay and user-chosen tunnels/VPNs only extend access. Copy Groups exports (zip) and imports durable state, CCCC_HOME backup/restore ports everything, formats are zip/JSONL/YAML/Markdown; account deletion cuts devices and removes the record while uninstalls intentionally keep data.
- Workspace files and live state do not port; copies exclude secrets.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Docker deployment documented: build image from source, run container, verify; image includes Claude Code, Codex CLI and Factory CLI.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
cccc:OWN-03: Code, prompts, credentials, logs and artifacts live on the user machine under CCCC_HOME (owner-only permissions) and are never uploaded by the product; Connect/Reach relay only messages/files between the user machines, and the account service stores just identities, device/host records and connection metadata with 180-day handled-request deletion. No telemetry client or statement, training-use, residency or subprocessor statements were found; provider separation holds via BYO runtime logins (CCCC even disables some provider telemetry flags). cross-checked
cccc:OWN-03
Code, prompts, credentials, logs and artifacts live on the user machine under CCCC_HOME (owner-only permissions) and are never uploaded by the product; Connect/Reach relay only messages/files between the user machines, and the account service stores just identities, device/host records and connection metadata with 180-day handled-request deletion. No telemetry client or statement, training-use, residency or subprocessor statements were found; provider separation holds via BYO runtime logins (CCCC even disables some provider telemetry flags).
- Local-first is structurally evidenced; business/privacy attestations are absent.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Single Rust daemon owns groups/ledger/tasks/context/lifecycle; ledger.jsonl append-only under CCCC_HOME/groups/<id>/; foreman coordinates + works, peers manage only self; MCP capability groups; React+TS+Vite+Tailwind+xterm Web, Rust+Axum host.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: Not a DAG workflow orchestrator; requirements Linux x86-64 glibc 2.28+/Apple Silicon macOS 11+/Windows x86-64 + agent CLI; 16 named CLIs + custom; read receipts via inbox reads; phone via tunnel/IM/Tailscale; no worktree cleanup by product.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Telegram/Slack/Discord/Mattermost/Feishu-DingTalk/WeCom/Weixin bridges: token setup, /send /status /pause /resume, subscribe authorization; Weixin direct-bot-chats only.
-
Finding: Native terminal output uses bounded memory with opt-in bounded per-agent transcript persistence; restart/shutdown boundaries documented.
cccc:OWN-04: CCCC has no paid plan and no prices: free Apache-2.0 download (install script, pip wheel, Docker, source) with no currency, intervals, seats, overages, taxes or credits anywhere in the checked sources. Agent spend is the user own existing subscriptions or API accounts, handled provider-side. cross-checked
cccc:OWN-04
CCCC has no paid plan and no prices: free Apache-2.0 download (install script, pip wheel, Docker, source) with no currency, intervals, seats, overages, taxes or credits anywhere in the checked sources. Agent spend is the user own existing subscriptions or API accounts, handled provider-side.
- Priced dimensions are absent because no paid plan exists; no purchase or billing test was run.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Connect existing native sessions
Finding: User-run Codex/Claude/Pi/OpenCode V2 sessions join Groups via enable + connect; CCCC imports no history, needs no provider key; experimental, tested Linux/WSL, macOS/native Windows need acceptance testing; Antigravity cannot connect this way.
cccc:OWN-05: No scenario can be priced in-product because the product itself costs $0 and all variable spend is provider-side and unstated: one, two and five developers run the same free download with their own runtime accounts, concurrency/token/compute assumptions unstated. Fixed cost is $0; unknown cost is whatever the chosen providers charge. cross-checked
cccc:OWN-05
No scenario can be priced in-product because the product itself costs $0 and all variable spend is provider-side and unstated: one, two and five developers run the same free download with their own runtime accounts, concurrency/token/compute assumptions unstated. Fixed cost is $0; unknown cost is whatever the chosen providers charge.
- Only the $0 product cost is statable; workload costs are provider-side.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.
cccc:OWN-06: Stable-channel product (0.4.55, Production/Stable classifier, near-daily releases) with public installers and no access gates; support runs through GitHub issues, Discord and Telegram with diagnostics tooling, and security reports get 3-business-day acknowledgment/7-day triage. No status page exists; no shutdown/deprecation is announced; export runs through Copy Groups and CCCC_HOME backup. cross-checked
cccc:OWN-06
Stable-channel product (0.4.55, Production/Stable classifier, near-daily releases) with public installers and no access gates; support runs through GitHub issues, Discord and Telegram with diagnostics tooling, and security reports get 3-business-day acknowledgment/7-day triage. No status page exists; no shutdown/deprecation is announced; export runs through Copy Groups and CCCC_HOME backup.
- Support/status facts are gaps; deprecation absence is not a finding.
-
Finding: Latest stable release v0.4.55 published 2026-11-12, with near-daily prior releases (v0.4.50–v0.4.54 within the preceding week).
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Daemon recovery ladder (actor restart, group stop/start, daemon stop/start), health baseline (doctor, daemon status, groups), CCCC_HOME backup/restore, upgrade playbook, Mail advisory semantics.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Send/Send+Reply/Mail contracts, consuming inbox reads, runtime.delivery facts, in-place question answers, 100 pins per group, push notification semantics, automation triggers/actions, per-actor secrets, Copy Groups export/import.
-
Finding: v0.4.55 adds project backlog Markdown file with Tasks|Backlog switch, PATH-change resume, notes removal, blocked-task fix (#201), plus upgrade caveats (in-page drafts/unsent edits live only in open page).
-
Finding: Distribution name cccc-pair 0.4.55, Apache-2.0, author ChesterRa, Development Status 5 Production/Stable, console environment, Linux/macOS/Windows classifiers, repo/issue URLs.
-
Finding: Security fixes prioritized for latest RC + stable; private advisories or maintainer email; acknowledgment within 3 business days, triage within 7.
-
Finding: Support via GitHub issues with version/OS/command/repro template, cccc doctor diagnostics, Discord and Telegram community channels.
cccc:OWN-08: The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist. cross-checked
cccc:OWN-08
The free Apache-2.0 download/installer/Docker/source is the ongoing free tier with no paid plan, trial, credits, card or gates: phone access, BYO subscriptions, Connect collaboration and review all work free. Self-hosting costs are the user own hardware plus provider accounts; no usage caps beyond incidental product limits exist.
- Free means the whole product; provider spend stays provider-side.
-
Bounded no-implementation search (pinned release)
Finding: Bounded search at the pinned release found no implementation or statement for: product telemetry clients or telemetry/consent UI (telemetry/analytics/posthog/sentry hits are provider flags CCCC disables, not product telemetry); SSO/SAML/OIDC/SCIM; editor/IDE extensions (vscode/vsix); local-model runtimes (ollama/lmstudio/llama.cpp); product PR/merge/release/deploy mechanics (gh pr/pull request hits are chat examples); spend/token/turn/time limits or cost tracking (budget/max_tokens/rate_limit hits are unrelated); concurrency quotas or CPU/RAM scheduling (concurrency/cpu_limit/memory_limit hits are dispatch locks); third-party MCP consumption by CCCC itself; OS/process/network sandboxing for agents; webhooks (only provider-side routine/task events); audit export; issue/label/epic entities beyond tasks; workspace export/import beyond Copy Groups; pricing, plans, seats, trials, credits, billing or status pages. Search terms: telemetry, analytics, sentry, sso, saml, oidc, scim, vscode, vsix, ollama, local model, pull request, spend_limit, token_limit, budget, max_tokens, concurrency, cpu_limit, memory_limit, webhook, sandbox, container, microvm, audit, kanban, export, pricing, billing, status page.
-
CCCC — Coordinate your coding agents like a group chat
Finding: Local-first open-source group-chat orchestrator for Claude Code/Codex/etc: shared Group with ledger, Send/Send+Reply/Mail, foreman+peers, Web/terminal/CLI/MCP/IM sharing daemon state, Chet assistant, Connect across instances, phone remote ops.
-
Finding: Local CCCC works without an account; remote access publishes a public hostname. Stores sign-in identities, device names/hostnames, Group connections, private requests/replies; handled requests deleted 180 days after last update; account deletion removes the record. Preview remote access supports Linux and macOS; capacity/availability not guaranteed.
-
Finding: Install via shell installer, PowerShell, or pip wheel (native executable in platform wheel); platforms Linux x86-64 glibc 2.28+, Apple Silicon macOS 11+, Windows x86-64. cccc starts daemon+Web UI (+status panel); headless mode; Admin Access Token before non-local exposure; managed Remote Access via pinned cloudflared.
-
Finding: Mobile-first React control plane: Messages/Terminals views, @mentions, reply/forward/pin, task board with Planned/Active/Done + backlog, AGENTS.md/Agent State/Skills panels, file browser with Git Changes (read-only diffs), Web Push notifications, automation rules, IM bridge config, LAN/tunnel/Tailscale remote access, Admin token security model.
-
Finding: Named runtimes: claude, codex, grok, opencode, kilo, copilot, cursor, devin, kiro, antigravity, droid, auggie, hermes, kimi, cline, amp, pi, deepseek, web_model, chatgpt_dots, grok_web_model, custom. Many start with approval prompts bypassed; managed Claude default bg worktree isolation; CCCC does not create/manage worktrees.
-
Finding: Full cccc CLI: daemon lifecycle, attach/groups/group/use, actor add/start/stop/restart/new-session/remove/secrets, send/tracked-send/reply/deliver/cancel-reply/inbox/inbox-summary/tail/ledger search, im, space, setup, update, web, mcp; CCCC_HOME/WEB_HOST/WEB_PORT env vars.
-
Finding: Connect links selected Groups across instances (same account, member invite, or direct no-account pairing); carries messages/replies/small files only, not terminals/workspace; each instance keeps its own state.
-
Finding: Documented patterns: solo agent, builder+reviewer pair, multi-agent team, phone monitoring via tunnel/IM, overnight tasks with automation; reviewer agent pattern for validation.
-
Finding: Root LICENSE is the Apache License, Version 2.0 grant covering the repository contents including daemon, Web UI, CLI and docs.