ai-devkit
Open-source Mac and Linux app. Agents run on your computer or on your own server.
Product websitegithub.com Compare with others Missing or wrong data?
- Desktop app
- Mac and Linux
- Browser
- No
- Phone
- No
- Command line
- Yes
- Agents
- Claude Code, Codex, Gemini CLI, 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
- No
- Coordinator agent
- Yes
- Built-in issues
- Yes
- Pull requests
- No
- Review screen
- No
- Team
- Not listed
- 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
- · oldest answer 2026-11-08
- Findings
- 35 published · 25 cross-checked by a second agent · read them all
- Sources
- 24 public pages on 2 sites: github.com, ai-devkit.com
- Screenshot
- raw.githubusercontent.com, retrieved · first party repository · shows appearance, not capability
- Data
- /products/ai-devkit.json · .md
- Method
- Same rules for all 103 researched ADEs. How we check · Report wrong data
What it does
Where you use it
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
Source and scope for Where you use itAgents it runs
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
Source and scope for Agents it runsAccounts and subscriptions
Underlying agent logins/provider credentials supply access. Capacity reads credentials already stored by provider tools.
Source and scope for Accounts and subscriptionsWhere agents run
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
Source and scope for Where agents runHow work is kept apart
Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
Source and scope for How work is kept apartParallel work
Several agents can run in separate managed sessions and be supervised together.
Source and scope for Parallel workScheduled runs
The scoped CLI/daemon and official plugin set has no schedule or external-event service that starts agent runs. A daemon discovery sweep only refreshes session state.
Source and scope for Scheduled runsGit and pull requests
Shipped guides can instruct an agent to use Git and manually created worktrees. The scoped CLI/plugin implementation has no product-owned PR creation or management workflow.
Source and scope for Git and pull requestsLicence and source
AI DevKit application source is Apache-2.0 licensed.
Source and scope for Licence and sourcePrice
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
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 |
| Desktop OS: Windows | No |
| Desktop OS: Linux | Yes |
| Web client | No |
| Phone access: Native iOS | No |
| Phone access: Native Android | No |
| Phone access: Phone browser | No |
| CLI | Yes |
| Coding agents: Claude Code | Yes |
| Coding agents: Codex | Yes |
| Coding agents: Cursor Agent | No |
| Coding agents: Gemini CLI | Yes |
| Coding agents: OpenCode | Yes |
| Coding agents: Pi | Yes |
| Coding agents: Grok | Yes |
| Coding agents: Aider | No |
| Coding agents: Antigravity | Yes |
| Use my subscription | Yes cross-checked |
| Where agents run: On your computer | Yes |
| Where agents run: On your own server | Yes |
| Where agents run: In the vendor's cloud | No |
| Workspace isolation: Git worktree | Yes |
| Workspace isolation: Container | No |
| Workspace isolation: VM or microVM | No |
| Parallel runs | Yes cross-checked |
| Coordinator | Yes cross-checked |
| Native issue tracking | Yes |
| Unattended triggers | No |
| Shared prompting | Unknown |
| Team roles | No |
| SSO | No |
| Pull request workflow | No |
| Review interface | No |
| 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 |
| Pricing model: Free plan | No |
| Pricing model: Subscription | No |
| Pricing model: Per seat | No |
| Pricing model: Usage-based | No |
| Pricing model: Per machine or compute | No |
| Pricing model: Bring your own key | Yes |
| Pricing model: One-time purchase | No |
| 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 |
| Open-source license | Yes. Apache-2.0 cross-checked |
| Any desktop app | Yes |
| Paid plan up to $20/month | Unknown |
| Paid plan up to $40/month | Unknown |
| Any phone client | No |
Filter receipts
Desktop OS: macOS: Yes
Desktop OS: macOS
Typed value: Yes.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Desktop OS: Windows: No
Desktop OS: Windows
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Desktop OS: Linux: Yes
Desktop OS: Linux
Typed value: Yes.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Web client: No
Web client
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Phone access: Native iOS: No
Phone access: Native iOS
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Phone access: Native Android: No
Phone access: Native Android
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Phone access: Phone browser: No
Phone access: Phone browser
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
CLI: Yes
CLI
Typed value: Yes.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Cli: ai-devkit commands and interactive console.
- Task actions: Console start/rename/kill/message and terminal focus.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Claude Code: Yes
Coding agents: Claude Code
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Codex: Yes
Coding agents: Codex
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Cursor Agent: No
Coding agents: Cursor Agent
Typed value: No.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Custom adapters: Managed start selects the shipped fixed runtime profiles; Cursor/Aider are not startable profiles in this revision.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
ai-devkit packages/agent-manager/src/harnesses/runtimeProfiles.ts
Finding: The complete startable runtime registry includes Claude, Codex, Copilot, Gemini CLI, Grok CLI, Antigravity CLI, OpenCode, Pi, Kiro, Devin and Muse. Cursor and Aider have no startable runtime profile or coding-harness adapter; configuring editor files is separate.
Coding agents: Gemini CLI: Yes
Coding agents: Gemini CLI
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: OpenCode: Yes
Coding agents: OpenCode
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Pi: Yes
Coding agents: Pi
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Grok: Yes
Coding agents: Grok
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Session integration: Reads live/historical conversation files; durable mode limited to Claude/Codex/Pi.
- Mixed harness operation: Multiple named provider sessions in console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
Coding agents: Aider: No
Coding agents: Aider
Typed value: No.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Custom adapters: Managed start selects the shipped fixed runtime profiles; Cursor/Aider are not startable profiles in this revision.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
ai-devkit packages/agent-manager/src/harnesses/runtimeProfiles.ts
Finding: The complete startable runtime registry includes Claude, Codex, Copilot, Gemini CLI, Grok CLI, Antigravity CLI, OpenCode, Pi, Kiro, Devin and Muse. Cursor and Aider have no startable runtime profile or coding-harness adapter; configuring editor files is separate.
Coding agents: Antigravity: Yes
Coding agents: Antigravity
Typed value: Yes.
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Harnesses: Claude Code, Codex, Gemini CLI, OpenCode, Pi, Grok CLI, Copilot, Kiro, Devin, Muse and Antigravity CLI. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
ai-devkit Antigravity support at cli@0.73.1
Finding: AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
-
ai-devkit Antigravity support at cli@0.73.1
Finding: AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
Use my subscription: Yes cross-checked
Use my subscription
Typed value: Yes.
Underlying agent logins/provider credentials supply access. Capacity reads credentials already stored by provider tools.
- Agent subscriptions: Logged-in provider sessions and quota readings reused; entitlement remains provider-owned.
- Api accounts: Agent/provider configuration determines API credentials.
- Vendor account requirement: Local app, not hosted product account.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/15-capacity.md
Finding: Readiness checks are hints, not successful login probes. Capacity reads stored provider credentials and reports quota windows, availability and reset times; it does not establish task billing.
Where agents run: On your computer: Yes
Where agents run: On your computer
Typed value: Yes.
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
- Local: Local processes and tmux sessions.
- Managed cloud: README explicitly not a hosted service and no remote cloud agent execution.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
Where agents run: On your own server: Yes
Where agents run: On your own server
Typed value: Yes.
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
- Byo remote: Install the same core CLI/daemon on a user-controlled Linux host; its systemd persistence and tmux execution survive a controlling SSH client logout.
- Client disconnect: Daemon/systemd and tmux sessions remain on the execution host when the remote control client disconnects, provided that host remains running.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
Where agents run: In the vendor's cloud: No
Where agents run: In the vendor's cloud
Typed value: No.
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
- Local: Local processes and tmux sessions.
- Managed cloud: README explicitly not a hosted service and no remote cloud agent execution.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
Workspace isolation: Git worktree: Yes
Workspace isolation: Git worktree
Typed value: Yes.
Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
- Security boundary: Permission guide warns presets grant broad shell/file access.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
Workspace isolation: Container: No
Workspace isolation: Container
Typed value: No.
Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
- Worktree: User runs git worktree add manually and supplies the directory; the ADE does not provision the worktree itself.
- Container: Complete runtime bundle is tmux/Herdr without a product container sandbox backend.
- Vm: No VM/microVM isolation backend in the scoped runtime bundle.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
Workspace isolation: VM or microVM: No
Workspace isolation: VM or microVM
Typed value: No.
Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
- Worktree: User runs git worktree add manually and supplies the directory; the ADE does not provision the worktree itself.
- Container: Complete runtime bundle is tmux/Herdr without a product container sandbox backend.
- Vm: No VM/microVM isolation backend in the scoped runtime bundle.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
Parallel runs: Yes cross-checked
Parallel runs
Typed value: Yes.
Several agents can run in separate managed sessions and be supervised together.
- Concurrent sessions: Managed workers in separate tmux sessions.
- Concurrent tasks: Lead can assign separate worker tasks.
- Status visibility: Console list and transcript preview.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
Coordinator: Yes cross-checked
Coordinator
Typed value: Yes.
AI DevKit orchestration skills equip a lead to start, message, inspect and wait for workers across harnesses.
- Roles: Lead and worker pattern through installed orchestration skills.
- Messages: agent send --id/--group and --stdin.
- Results return: send --wait reads worker responses.
- Cross harness: Claude lead with Codex/OpenCode worker examples.
- Orchestration owner: AI DevKit command and skill layer; underlying harness executes.
- Live progress: agent list/detail/console.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
Native issue tracking: Yes
Native issue tracking
Typed value: Yes.
The optional first-party task-manager plugin provides persistent independent tasks, statuses, phases, blockers, evidence, artifacts, ownership and event history through ai-devkit task.
- Create: ai-devkit task create creates a durable task record.
- Edit: update/status/phase/progress commands revise task state.
- Task session separation: tasks.db has stable task IDs and event rows; session attribution is optional metadata.
- Metadata: Tasks retain title, name, phase, owner, blockers, evidence and branch/PR links.
- Durable states: Status and phase persist in SQLite snapshots and events.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit packages/task-manager/README.md
Finding: The official installable task-manager plugin adds ai-devkit task create/list/show/update/status/progress/assign/note/close commands. Stable independent task IDs, metadata and append-only events persist in tasks.db, separate from agent sessions.
Unattended triggers: No
Unattended triggers
Typed value: No.
The scoped CLI/daemon and official plugin set has no schedule or external-event service that starts agent runs. A daemon discovery sweep only refreshes session state.
- Schedules: No product schedule registration/start service in the complete scoped command/RPC/plugin set.
- Events: Chat bridge messages steer an already connected agent; no external-event launcher is implemented.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
Team roles: No
Team roles
Typed value: No.
The scoped application uses local OS-account state and chat-service bridge authorization. It has no named human workspace roles or enterprise SSO.
- Organization roles: No human workspace/organization role or membership model in the scoped app.
- Sso: No enterprise SSO path in the local RPC/CLI/bridge configuration.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
SSO: No
SSO
Typed value: No.
The scoped application uses local OS-account state and chat-service bridge authorization. It has no named human workspace roles or enterprise SSO.
- Organization roles: No human workspace/organization role or membership model in the scoped app.
- Sso: No enterprise SSO path in the local RPC/CLI/bridge configuration.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
Pull request workflow: No
Pull request workflow
Typed value: No.
Shipped guides can instruct an agent to use Git and manually created worktrees. The scoped CLI/plugin implementation has no product-owned PR creation or management workflow.
- Pull requests: Task branch/PR links are metadata; no PR creation/management CLI or RPC is supplied.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
Review interface: No
Review interface
Typed value: No.
Phase Markdown and historical transcripts remain available locally; output can be piped into agent send. The product console has no dedicated code-change diff review view.
- Diffs: Console renders transcript Markdown and memory; no product-owned proposed-code diff viewer in the scoped client.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
Product license: Apache-2.0: Yes. Apache-2.0 cross-checked
Product license: Apache-2.0
Typed value: Yes. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: MIT: No. Apache-2.0 cross-checked
Product license: MIT
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: GPL-3.0: No. Apache-2.0 cross-checked
Product license: GPL-3.0
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
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.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: AGPL-3.0: No. Apache-2.0 cross-checked
Product license: AGPL-3.0
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
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.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: Elastic License 2.0: No. Apache-2.0 cross-checked
Product license: Elastic License 2.0
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: Custom terms: No. Apache-2.0 cross-checked
Product license: Custom terms
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Product license: Proprietary: No. Apache-2.0 cross-checked
Product license: Proprietary
Typed value: No. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Self-hosting: Yes cross-checked
Self-hosting
Typed value: Yes.
Core CLI/daemon, memory and console run on user-controlled machine; no hosted service required.
- Self hosting: Local npm CLI and Rust daemon.
- Formats: JSON config, SQLite memory and agent-owned session files.
- Account independence: No hosted AI DevKit account.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
Free access: Yes cross-checked
Free access
Typed value: Yes.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: Free: Yes. Free cross-checked
Starting price: Free
Typed value: Yes. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: Up to $20/month: No. Free cross-checked
Starting price: Up to $20/month
Typed value: No. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: $21–40/month: No. Free cross-checked
Starting price: $21–40/month
Typed value: No. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: $41–100/month: No. Free cross-checked
Starting price: $41–100/month
Typed value: No. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: Over $100/month: No. Free cross-checked
Starting price: Over $100/month
Typed value: No. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Starting price: Contact for pricing: No. Free cross-checked
Starting price: Contact for pricing
Typed value: No. Free.
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Ongoing free: The ai-devkit package is free to run through npx under Apache-2.0.
- Self host cost: Runs on the user's machine with no AI DevKit fee; agent provider charges remain separate.
- Paid only: Not paid-only: the package runs free of charge from npm.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.
Pricing model: Free, no paid plan: Yes
Pricing model: Free, no paid plan
Typed value: Yes.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Free plan: No
Pricing model: Free plan
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Subscription: No
Pricing model: Subscription
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Per seat: No
Pricing model: Per seat
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Usage-based: No
Pricing model: Usage-based
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Per machine or compute: No
Pricing model: Per machine or compute
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Pricing model: Bring your own key: Yes
Pricing model: Bring your own key
Typed value: Yes.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Model cost: The official per-harness setup guide explicitly supports a user-supplied Google API key for managed Gemini CLI workers. This is a provider BYOK model-use path, separate from the free ADE application. Provider rates are separate from the ADE fee.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
AI DevKit per-harness authentication setup
Finding: The official ai-devkit orchestration guide explicitly lists Google API-key authentication as an alternative to login for the managed Gemini CLI worker. The application orchestrates agents authenticated to users’ own providers; it does not supply a bundled model quota.
Pricing model: One-time purchase: No
Pricing model: One-time purchase
Typed value: No.
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- Compute cost: Runs locally and does not run a remote cloud agent, so there is no vendor compute charge described.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
Keeps going with laptop closed: Yes
Keeps going with laptop closed
Typed value: Yes.
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
- Byo remote: Install the same core CLI/daemon on a user-controlled Linux host; its systemd persistence and tmux execution survive a controlling SSH client logout.
- Client disconnect: Daemon/systemd and tmux sessions remain on the execution host when the remote control client disconnects, provided that host remains running.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
Open-source license: Yes. Apache-2.0 cross-checked
Open-source license
Typed value: Yes. Apache-2.0.
AI DevKit application source is Apache-2.0 licensed.
- Client license: Apache-2.0.
- Source visibility: Public application source.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
Any desktop app: Yes
Any desktop app
Typed value: Yes.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
Any phone client: No
Any phone client
Typed value: No.
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Desktop os: macOS/Linux managed terminal interface; Windows console unsupported.
- Mobile apps: No native iOS/Android client in the scoped application build targets.
- Browser: The scoped product contains CLI/TUI and documentation website, without a browser control client.
- Mobile browser: Chat-service bridges do not provide a product-owned phone-browser client; no such client is built in the scoped tree.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
All public findings
Findings may summarize a source. They are not presented as verbatim statements.
ai-devkit:ACC-01: AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
ai-devkit:ACC-01
AI DevKit has native macOS/Linux CLI/TUI operation; the managed console is unsupported on Windows. The scoped clients are terminal interfaces, with separate Telegram/Slack bridges rather than product web or phone apps.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
-
ai-devkit web/content/docs/17-multi-agent-orchestration.md
Finding: Managed agent start and experimental terminal console support macOS/Linux and explicitly do not support Windows. The scoped product client tree contains CLI/TUI, without browser, native iOS/Android or phone-browser control clients; its web directory is a documentation site and chat-service bridges are separate.
ai-devkit:ACC-02: Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
ai-devkit:ACC-02
Managed start supports named coding harnesses and supervision across providers; configuration integration alone is not proof that every editor is launchable. AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy. The complete startable runtime registry omits Cursor and Aider.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
- Antigravity CLI must be installed and authenticated; editor configuration support alone does not establish a startable agent.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
ai-devkit Antigravity support at cli@0.73.1
Finding: AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
-
ai-devkit Antigravity support at cli@0.73.1
Finding: AI DevKit cli@0.73.1 registers AntigravityCliAdapter and a startable antigravity_cli runtime whose command is agy.
-
ai-devkit packages/agent-manager/src/harnesses/runtimeProfiles.ts
Finding: The complete startable runtime registry includes Claude, Codex, Copilot, Gemini CLI, Grok CLI, Antigravity CLI, OpenCode, Pi, Kiro, Devin and Muse. Cursor and Aider have no startable runtime profile or coding-harness adapter; configuring editor files is separate.
ai-devkit:ACC-03: Underlying agent logins/provider credentials supply access. Capacity reads credentials already stored by provider tools. cross-checked
ai-devkit:ACC-03
Underlying agent logins/provider credentials supply access. Capacity reads credentials already stored by provider tools.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/15-capacity.md
Finding: Readiness checks are hints, not successful login probes. Capacity reads stored provider credentials and reports quota windows, availability and reset times; it does not establish task billing.
ai-devkit:ACC-04: Machine setup installs skills/hooks/memory wiring; init writes project config. Console needs interactive TTY and installed agent runtimes. cross-checked
ai-devkit:ACC-04
Machine setup installs skills/hooks/memory wiring; init writes project config. Console needs interactive TTY and installed agent runtimes.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
ai-devkit:ACC-05: CLI exposes managed agent control, memory and daemon operations; memory MCP server provides tool access. cross-checked
ai-devkit:ACC-05
CLI exposes managed agent control, memory and daemon operations; memory MCP server provides tool access.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
Finding: Local SQLite stores titled/tagged project/global memory. Memory MCP server and CLI provide retrieval/storage; setup wires compatible agents, Pi uses CLI/skill instead.
ai-devkit:ACC-06: Unified config reconciles per-agent instructions/skills/MCP setup; compatible agents receive memory-server wiring. cross-checked
ai-devkit:ACC-06
Unified config reconciles per-agent instructions/skills/MCP setup; compatible agents receive memory-server wiring.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
Finding: Local SQLite stores titled/tagged project/global memory. Memory MCP server and CLI provide retrieval/storage; setup wires compatible agents, Pi uses CLI/skill instead.
ai-devkit:PLAN-02: Lifecycle skill writes requirements/design/planning/implementation/testing files under docs/ai and requires staged review by instruction. cross-checked
ai-devkit:PLAN-02
Lifecycle skill writes requirements/design/planning/implementation/testing files under docs/ai and requires staged review by instruction.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
ai-devkit:PLAN-04: Local SQLite memory supports titled/tagged/scope records through MCP and CLI, retrieved when relevant. cross-checked
ai-devkit:PLAN-04
Local SQLite memory supports titled/tagged/scope records through MCP and CLI, retrieved when relevant.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local SQLite stores titled/tagged project/global memory. Memory MCP server and CLI provide retrieval/storage; setup wires compatible agents, Pi uses CLI/skill instead.
ai-devkit:PLAN-05: Memory retrieval supplies prior project knowledge on demand. This is not a full code index. cross-checked
ai-devkit:PLAN-05
Memory retrieval supplies prior project knowledge on demand. This is not a full code index.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local SQLite stores titled/tagged project/global memory. Memory MCP server and CLI provide retrieval/storage; setup wires compatible agents, Pi uses CLI/skill instead.
ai-devkit:PLAN-06: The optional first-party task-manager plugin provides persistent independent tasks, statuses, phases, blockers, evidence, artifacts, ownership and event history through ai-devkit task.
ai-devkit:PLAN-06
The optional first-party task-manager plugin provides persistent independent tasks, statuses, phases, blockers, evidence, artifacts, ownership and event history through ai-devkit task.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit packages/task-manager/README.md
Finding: The official installable task-manager plugin adds ai-devkit task create/list/show/update/status/progress/assign/note/close commands. Stable independent task IDs, metadata and append-only events persist in tasks.db, separate from agent sessions.
ai-devkit:PLAN-08: Lifecycle skill persists requirements/design/plans in repo Markdown; no server-enforced spec-drift gate established. cross-checked
ai-devkit:PLAN-08
Lifecycle skill persists requirements/design/plans in repo Markdown; no server-enforced spec-drift gate established.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
ai-devkit:COORD-01: Several agents can run in separate managed sessions and be supervised together. cross-checked
ai-devkit:COORD-01
Several agents can run in separate managed sessions and be supervised together.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
ai-devkit:COORD-02: Lead uses named workers and start-then-assign handshake; same-folder session detection collisions are documented. cross-checked
ai-devkit:COORD-02
Lead uses named workers and start-then-assign handshake; same-folder session detection collisions are documented.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
ai-devkit:COORD-03: AI DevKit orchestration skills equip a lead to start, message, inspect and wait for workers across harnesses. cross-checked
ai-devkit:COORD-03
AI DevKit orchestration skills equip a lead to start, message, inspect and wait for workers across harnesses.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
ai-devkit:COORD-04: The scoped CLI/daemon and official plugin set has no schedule or external-event service that starts agent runs. A daemon discovery sweep only refreshes session state.
ai-devkit:COORD-04
The scoped CLI/daemon and official plugin set has no schedule or external-event service that starts agent runs. A daemon discovery sweep only refreshes session state.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
ai-devkit:COORD-05: Console messages or Telegram/Slack bridge can steer running agents; managed kill requires confirmation. cross-checked
ai-devkit:COORD-05
Console messages or Telegram/Slack bridge can steer running agents; managed kill requires confirmation.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
web/content/docs/12-channel.md
Finding: Bridge forwards messages to running terminal agents and polls output. Telegram restricts each session to first user. Slack proof of concept allows workspace members who can DM the app to send to same connected agent; responses route to first DM until restart.
ai-devkit:COORD-07: Composable lifecycle/verification/orchestration skills guide workflows without prescribing one fixed agent. cross-checked
ai-devkit:COORD-07
Composable lifecycle/verification/orchestration skills guide workflows without prescribing one fixed agent.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
ai-devkit:RUN-01: The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
ai-devkit:RUN-01
The local CLI/daemon runs agents on its own macOS/Linux host, including a user-controlled always-on Linux server reached by SSH. systemd user persistence survives logout; this does not keep a sleeping execution host running.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/16-runtime.md
Finding: Rust devkitd auto-spawns over Unix socket, logs locally and supports Linux systemd user persistence; daemon registry/persistence is local.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
ai-devkit:RUN-02: Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
ai-devkit:RUN-02
Worktree setup is manual guidance: users create directories and pass --cwd. The scoped runtime bundle supplies terminal processes without product-owned container or VM isolation.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
-
web/content/docs/19-permissions.md
Finding: Linux 0.70.1 guidance configures underlying harness allow-lists/trust, with broad shell/file permissions. Other harnesses need verification. This is not an AI DevKit security sandbox.
-
ai-devkit rust/crates/devkitd/src/runtime.rs
Finding: The complete runtime bundle launches agents through tmux or Herdr on the daemon host. The local Unix-socket control plane can run on a user-controlled Linux server and systemd keeps it alive after logout; no vendor service is required. It has no automatic worktree provisioning, container isolation or VM runtime.
ai-devkit:RUN-03: Managed durable mode is limited to Claude/Codex/Pi; daemon persistence can be installed on Linux. No full worker crash/reboot recovery test was run. cross-checked
ai-devkit:RUN-03
Managed durable mode is limited to Claude/Codex/Pi; daemon persistence can be installed on Linux. No full worker crash/reboot recovery test was run.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
web/content/docs/16-runtime.md
Finding: Rust devkitd auto-spawns over Unix socket, logs locally and supports Linux systemd user persistence; daemon registry/persistence is local.
ai-devkit:RUN-05: send --wait supports timeout and readiness handshake; quotas are readings, not enforced spend limits. cross-checked
ai-devkit:RUN-05
send --wait supports timeout and readiness handshake; quotas are readings, not enforced spend limits.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
-
web/content/docs/15-capacity.md
Finding: Readiness checks are hints, not successful login probes. Capacity reads stored provider credentials and reports quota windows, availability and reset times; it does not establish task billing.
ai-devkit:RUN-06: Console shows last-observed running/waiting/idle/unknown status and transcripts; status readiness is explicitly a hint. cross-checked
ai-devkit:RUN-06
Console shows last-observed running/waiting/idle/unknown status and transcripts; status readiness is explicitly a hint.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
web/content/docs/15-capacity.md
Finding: Readiness checks are hints, not successful login probes. Capacity reads stored provider credentials and reports quota windows, availability and reset times; it does not establish task billing.
-
web/content/docs/16-runtime.md
Finding: Rust devkitd auto-spawns over Unix socket, logs locally and supports Linux systemd user persistence; daemon registry/persistence is local.
ai-devkit:RUN-07: Capacity reports provider quota windows and reset times. It does not establish per-task billed cost. cross-checked
ai-devkit:RUN-07
Capacity reports provider quota windows and reset times. It does not establish per-task billed cost.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/15-capacity.md
Finding: Readiness checks are hints, not successful login probes. Capacity reads stored provider credentials and reports quota windows, availability and reset times; it does not establish task billing.
ai-devkit:TEAM-01: Slack proof-of-concept bridge permits multiple workspace members to send text toward one agent; Telegram allows only first user per bridge session. cross-checked
ai-devkit:TEAM-01
Slack proof-of-concept bridge permits multiple workspace members to send text toward one agent; Telegram allows only first user per bridge session.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/12-channel.md
Finding: Bridge forwards messages to running terminal agents and polls output. Telegram restricts each session to first user. Slack proof of concept allows workspace members who can DM the app to send to same connected agent; responses route to first DM until restart.
ai-devkit:TEAM-02: The scoped application uses local OS-account state and chat-service bridge authorization. It has no named human workspace roles or enterprise SSO.
ai-devkit:TEAM-02
The scoped application uses local OS-account state and chat-service bridge authorization. It has no named human workspace roles or enterprise SSO.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
ai-devkit:TEAM-03: Underlying harness permissions/trust must be configured; tested presets are Linux 0.70.1 guidance with broad access. cross-checked
ai-devkit:TEAM-03
Underlying harness permissions/trust must be configured; tested presets are Linux 0.70.1 guidance with broad access.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/19-permissions.md
Finding: Linux 0.70.1 guidance configures underlying harness allow-lists/trust, with broad shell/file permissions. Other harnesses need verification. This is not an AI DevKit security sandbox.
ai-devkit:SHIP-01: Phase Markdown and historical transcripts remain available locally; output can be piped into agent send. The product console has no dedicated code-change diff review view.
ai-devkit:SHIP-01
Phase Markdown and historical transcripts remain available locally; output can be piped into agent send. The product console has no dedicated code-change diff review view.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/13-agent-console.md
Finding: Interactive TTY console lists agents, previews transcripts, starts/renames/kills agents and connects channels. Named managed start types include Claude, Codex, Gemini CLI, Grok CLI, OpenCode and Pi. Durable mode limited to Claude/Codex/Pi; managed runtime tmux or configured Herdr.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
ai-devkit:SHIP-02: verify/TDD/review skills instruct fresh verification before completion; no independent runtime-enforced gate was established. cross-checked
ai-devkit:SHIP-02
verify/TDD/review skills instruct fresh verification before completion; no independent runtime-enforced gate was established.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
ai-devkit:SHIP-03: Shipped guides can instruct an agent to use Git and manually created worktrees. The scoped CLI/plugin implementation has no product-owned PR creation or management workflow.
ai-devkit:SHIP-03
Shipped guides can instruct an agent to use Git and manually created worktrees. The scoped CLI/plugin implementation has no product-owned PR creation or management workflow.
- Source inspection covers the scoped release only. Unanswered atomic questions remain unknown; no live scenario was run.
-
ai-devkit rust/crates/devkitd/src/server.rs
Finding: The scoped daemon serves local Unix-socket agent RPCs; CLI commands and shipped plugins contain no schedule/webhook service that starts work, human membership/role or enterprise SSO model. Console preview renders conversation Markdown rather than a proposed-code diff viewer; Git/PR references in task records are metadata, without PR creation/management commands.
ai-devkit:OWN-01: AI DevKit application source is Apache-2.0 licensed. cross-checked
ai-devkit:OWN-01
AI DevKit application source is Apache-2.0 licensed.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Application repository source is Apache-2.0 licensed.
ai-devkit:OWN-02: Core CLI/daemon, memory and console run on user-controlled machine; no hosted service required. cross-checked
ai-devkit:OWN-02
Core CLI/daemon, memory and console run on user-controlled machine; no hosted service required.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/16-runtime.md
Finding: Rust devkitd auto-spawns over Unix socket, logs locally and supports Linux systemd user persistence; daemon registry/persistence is local.
ai-devkit:OWN-03: README explicitly says no telemetry and local memory. Channel tokens are local plaintext files mode 0600; agent/provider terms remain separate. cross-checked
ai-devkit:OWN-03
README explicitly says no telemetry and local memory. Channel tokens are local plaintext files mode 0600; agent/provider terms remain separate.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
Finding: Local-first control layer offers unified configuration, console, cross-agent messages, local SQLite memory and composable lifecycle/verification skills. It explicitly is not a hosted service, has no telemetry and does not run remote cloud agents.
-
web/content/docs/12-channel.md
Finding: Bridge forwards messages to running terminal agents and polls output. Telegram restricts each session to first user. Slack proof of concept allows workspace members who can DM the app to send to same connected agent; responses route to first DM until restart.
ai-devkit:OWN-04: AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
ai-devkit:OWN-04
AI DevKit is free (Apache-2.0) and not a hosted service; no paid plan found, and the agents/models it coordinates are paid separately to their providers. The managed Gemini CLI worker can use a user-supplied Google API key, so provider model usage can be paid through the user’s own key.
- First-party pricing pages, terms and licences read 2026-11-10; no purchase, billing or authenticated account test.
- Provider credentials and model usage are separate from the free ADE application; no billing test ran.
-
Finding: The site says AI DevKit is free and Apache-2.0 licensed.
-
Finding: The README says it is not a hosted service, runs locally and coordinates agents the user already uses.
-
Finding: The Apache-2.0 license grants a perpetual, no-charge, royalty-free copyright licence to use and redistribute the software.
-
AI DevKit per-harness authentication setup
Finding: The official ai-devkit orchestration guide explicitly lists Google API-key authentication as an alternative to login for the managed Gemini CLI worker. The application orchestrates agents authenticated to users’ own providers; it does not supply a bundled model quota.
ai-devkit:OWN-06: Latest listed CLI release 0.72.0 was published November 3; console remains experimental macOS/Linux and unsupported Windows. cross-checked
ai-devkit:OWN-06
Latest listed CLI release 0.72.0 was published November 3; console remains experimental macOS/Linux and unsupported Windows.
- Only the named facts are documented. Unanswered atomic questions remain unknown; no execution test was performed.
-
web/content/docs/17-multi-agent-orchestration.md
Finding: Product orchestration skill guides a lead using agent start/send/list to supervise named workers and cross-harness review. Users manually supply separate worktree directories; same-folder session matching may collide. Console experimental on macOS/Linux and unsupported Windows.
-
web/content/roadmap/10-agent-management.md
Finding: Agent management milestone complete. Future parity, ergonomics and platform hardening are not dated release promises.
-
Finding: 0.72.0 was published 2026-11-03. Release notes describe the scoped changes; this is not a hands-on check.
ai-devkit:OWN-08: AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers. cross-checked
ai-devkit:OWN-08
AI DevKit is free Apache-2.0 software run locally through npx, with no paid plan or hosted service. The coding agents it supervises are paid to their providers.
- Free scenarios, usage caps and card requirements for the underlying agents were not checked. No installed-product test was run.
-
Finding: The repository is licensed under the Apache License, Version 2.0.
-
AI DevKit README: setup and What it is not
Finding: AI DevKit is run free with npx ai-devkit@latest setup and init; the README says it is Apache-2.0, runs locally and is not a hosted service, and names no price or paid plan.