Zed
Open-source desktop app for Mac, Windows and Linux. Agents run on your computer or on your own server.
Product websitezed.dev Compare with others Missing or wrong data?
- Desktop app
- Mac, Windows and Linux
- Browser
- No
- Phone
- No
- Command line
- Yes
- Agents
- Claude Code, Codex, Cursor Agent, 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
- No
- Isolation
- Worktree and Container
- Parallel agents
- Yes
- Scheduled runs
- No
- Coordinator agent
- Yes
- Built-in issues
- No
- Pull requests
- No
- Review screen
- Yes
- Team
- Roles
- Self-hosting
- Yes
- Licence
- GPL-3.0-or-later
- Price from
- Free
- How it charges
- Free plan, subscription, per seat, usage-based and bring your own key
- Paid plans from
- $10 or less/month
Receipts
- Checked
- · oldest answer 2026-11-07
- Findings
- 40 published · 31 cross-checked by a second agent · read them all
- Sources
- 49 public pages on 3 sites: api.github.com, zed.dev, github.com
- GitHub stars
- 90,314 on · api.github.com · repository attention, not users or capability
- Screenshot
- zed.dev, retrieved · first party release · shows appearance, not capability
- Data
- /products/zed.json · .md
- Method
- Same rules for all 103 researched ADEs. How we check · Report wrong data
What it does
Where you use it
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
Source and scope for Where you use itAgents it runs
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
Source and scope for Agents it runsAccounts and subscriptions
A separate existing-subscription setup route is documented; support must be checked per provider and agent path. Zed Agent supports provider API access; provider-specific setup applies. Zed-hosted models use Zed account billing separately from own provider accounts. Local or self-hosted model endpoints are a documented Zed Agent route. Provider and external-agent authentication are configured separately; native Zed provider settings do not configure external-agent credentials. The editor itself needs no authentication; collaboration features require sign-in. ChatGPT Plus/Pro is a native subscription provider; Claude Pro/Max uses Claude external/CLI paths; OpenCode requires API keys and Cursor uses external/CLI paths.
Source and scope for Accounts and subscriptionsWhere agents run
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
Source and scope for Where agents runHow work is kept apart
A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
Source and scope for How work is kept apartParallel work
Multiple agent and terminal threads can run at once, against different projects. Multiple threads can receive independent tasks while previous threads continue. Docs permit multiple concurrent threads but give no numerical product concurrency quota. The Threads Sidebar groups independent conversations by project.
Source and scope for Parallel workScheduled runs
The native agent and agent UI start work from explicit session prompts. Source inspection found no product scheduler or external-event-to-agent run starter.
Source and scope for Scheduled runsGit and pull requests
The Git Panel reflects working-tree, staging and branch state, commits and undoes commits, manages branches and linked worktrees, and resolves merge conflicts. Its Create Pull Request action only opens the hosting provider's create page in the browser.
Source and scope for Git and pull requestsLicence and source
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
Source and scope for Licence and sourcePrice
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
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 | Yes |
| 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 | Yes |
| Coding agents: Gemini CLI | Yes |
| Coding agents: OpenCode | Yes |
| Coding agents: Pi | Yes |
| Coding agents: Grok | Yes |
| Coding agents: Aider | Unknown |
| 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 | Yes |
| Workspace isolation: VM or microVM | No |
| Parallel runs | Yes cross-checked |
| Coordinator | Yes cross-checked |
| Native issue tracking | No |
| Unattended triggers | No |
| Shared prompting | No |
| Team roles | Yes |
| SSO | No |
| Pull request workflow | No |
| Review interface | Yes cross-checked |
| Product license: Apache-2.0 | No. GPL-3.0-or-later cross-checked |
| Product license: MIT | No. GPL-3.0-or-later cross-checked |
| Product license: GPL-3.0 | No. GPL-3.0-or-later cross-checked |
| Product license: GPL-3.0-or-later | Yes. GPL-3.0-or-later cross-checked |
| Product license: AGPL-3.0 | No. GPL-3.0-or-later cross-checked |
| Product license: AGPL-3.0-or-later | No. GPL-3.0-or-later cross-checked |
| Product license: Elastic License 2.0 | No. GPL-3.0-or-later cross-checked |
| Product license: Custom terms | No. GPL-3.0-or-later cross-checked |
| Product license: Proprietary | No. GPL-3.0-or-later cross-checked |
| Self-hosting | Yes cross-checked |
| Telemetry opt-out | Yes cross-checked |
| 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 | No |
| Pricing model: Free plan | Yes |
| Pricing model: Subscription | Yes |
| Pricing model: Per seat | Yes |
| Pricing model: Usage-based | Yes |
| 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 | No. $10 or less/month |
| Paid plans from: $10 or less/month | Yes. $10 or less/month |
| Paid plans from: $11–20/month | No. $10 or less/month |
| Paid plans from: $21–40/month | No. $10 or less/month |
| Paid plans from: $41–100/month | No. $10 or less/month |
| Paid plans from: Over $100/month | No. $10 or less/month |
| Paid plans from: Contact sales | No. $10 or less/month |
| Lifecycle status: Active | Unknown |
| Lifecycle status: Sunsetting | Unknown |
| Keeps going with laptop closed | No |
| Open-source license | Yes. GPL-3.0-or-later cross-checked |
| Any desktop app | Yes |
| Paid plan up to $20/month | Yes. $10 or less/month |
| Paid plan up to $40/month | Yes. $10 or less/month |
| Any phone client | No |
Filter receipts
Desktop OS: macOS: Yes
Desktop OS: macOS
Typed value: Yes.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Desktop os: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides.
- Browser: The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Desktop OS: Windows: Yes
Desktop OS: Windows
Typed value: Yes.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Desktop os: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides.
- Browser: The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Desktop OS: Linux: Yes
Desktop OS: Linux
Typed value: Yes.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Desktop os: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides.
- Browser: The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Web client: No
Web client
Typed value: No.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Desktop os: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides.
- Browser: The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Phone access: Native iOS: No
Phone access: Native iOS
Typed value: No.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Mobile apps: The complete native installation list distributes desktop platforms only; no native iOS/Android build is distributed.
- Mobile browser: The browser version is explicitly not yet available; the current roadmap still places Zed on the Web upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed supported client platforms
Finding: The complete installation list is macOS, Linux and Windows. Other platforms are not yet available, and Web is explicitly listed as unavailable.
Phone access: Native Android: No
Phone access: Native Android
Typed value: No.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Mobile apps: The complete native installation list distributes desktop platforms only; no native iOS/Android build is distributed.
- Mobile browser: The browser version is explicitly not yet available; the current roadmap still places Zed on the Web upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed supported client platforms
Finding: The complete installation list is macOS, Linux and Windows. Other platforms are not yet available, and Web is explicitly listed as unavailable.
Phone access: Phone browser: No
Phone access: Phone browser
Typed value: No.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Mobile apps: The complete native installation list distributes desktop platforms only; no native iOS/Android build is distributed.
- Mobile browser: The browser version is explicitly not yet available; the current roadmap still places Zed on the Web upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed supported client platforms
Finding: The complete installation list is macOS, Linux and Windows. Other platforms are not yet available, and Web is explicitly listed as unavailable.
CLI: Yes
CLI
Typed value: Yes.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Cli: The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Product-owned zed CLI opens files/directories, remote projects and diff views; wait/reuse/add options integrate with other tools. This is an editor launcher, not a headless agent API.
Coding agents: Claude Code: Yes
Coding agents: Claude Code
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: Codex: Yes
Coding agents: Codex
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: Cursor Agent: Yes
Coding agents: Cursor Agent
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: Gemini CLI: Yes
Coding agents: Gemini CLI
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: OpenCode: Yes
Coding agents: OpenCode
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: Pi: Yes
Coding agents: Pi
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Custom adapters: Custom ACP agents configure an executable, arguments and environment through agent_servers.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
Coding agents: Grok: Yes
Coding agents: Grok
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Session integration: Registry installation adds the ACP agent to the Agent Panel/Threads Sidebar new-thread menu; Zed hosts its thread and the agent owns its runtime, auth and configuration.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Zed ACP Registry installation and agent threads
Finding: Zed installs registry agents through zed: acp registry or Agent Settings/Add Agent. Installed agents appear in the Agent Panel and Threads Sidebar new-thread menu. Zed manages their threads while each ACP agent owns its runtime and authentication; the common-agent list is explicitly not exhaustive.
-
Grok Build in the Zed ACP Registry
Finding: Zed provides a named Grok Build ACP Registry listing with install/select instructions and the native xAI CLI command npx @xai-official/grok agent stdio. This is the coding-agent runtime, not selecting a Grok model in Zed Agent.
Coding agents: Antigravity: Yes
Coding agents: Antigravity
Typed value: Yes.
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Harnesses: Official integrations cover Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside, plus named ACP Registry installs for Grok Build and Google Antigravity. The list is not exhaustive.
- Session integration: Registry installation adds the ACP agent to the Agent Panel/Threads Sidebar new-thread menu; Zed hosts its thread and the agent owns its runtime, auth and configuration.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Zed ACP Registry installation and agent threads
Finding: Zed installs registry agents through zed: acp registry or Agent Settings/Add Agent. Installed agents appear in the Agent Panel and Threads Sidebar new-thread menu. Zed manages their threads while each ACP agent owns its runtime and authentication; the common-agent list is explicitly not exhaustive.
-
Google Antigravity in the Zed ACP Registry
Finding: Zed provides a named Google Antigravity ACP Registry listing with a binary distribution and instructions to install/select it or configure its agy_acp_server.par command as an agent server.
Use my subscription: Yes cross-checked
Use my subscription
Typed value: Yes.
A separate existing-subscription setup route is documented; support must be checked per provider and agent path. Zed Agent supports provider API access; provider-specific setup applies. Zed-hosted models use Zed account billing separately from own provider accounts. Local or self-hosted model endpoints are a documented Zed Agent route. Provider and external-agent authentication are configured separately; native Zed provider settings do not configure external-agent credentials. The editor itself needs no authentication; collaboration features require sign-in. ChatGPT Plus/Pro is a native subscription provider; Claude Pro/Max uses Claude external/CLI paths; OpenCode requires API keys and Cursor uses external/CLI paths.
- Agent subscriptions: A separate existing-subscription setup route is documented; support must be checked per provider and agent path.
- Api accounts: Zed Agent supports provider API access; provider-specific setup applies.
- Vendor credits: Zed-hosted models use Zed account billing separately from own provider accounts.
- Local models: Local or self-hosted model endpoints are a documented Zed Agent route.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Native Zed AI has hosted, API, subscription, gateway and local model setup routes; these settings do not configure external agents.
Where agents run: On your computer: Yes
Where agents run: On your computer
Typed value: Yes.
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
- Local: Native agent terminal tools run shell commands in the project environment.
- Unknown after the targeted first-party documentation, release and repository pass: managed_cloud, hybrid, region, expiry, client_disconnect, other_device_continuation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
Where agents run: On your own server: Yes
Where agents run: On your own server
Typed value: Yes.
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
- Byo remote: SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established.
- Self hosted: Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH.
- Payer: The user supplies and pays for the remote server or devcontainer host.
- Resources: The user supplies local UI hardware and remote host compute.
- Unknown after the targeted first-party documentation, release and repository pass: managed_cloud, hybrid, region, expiry, client_disconnect, other_device_continuation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Local UI and model communication remain local while source files, language servers, tasks and terminals run on an SSH server.
Where agents run: In the vendor's cloud: No
Where agents run: In the vendor's cloud
Typed value: No.
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
- Managed cloud: No vendor execution mode in the native local-agent/SSH architecture; hosted model inference is separate.
- Client disconnect: The agent/model loop belongs to the controlling local app even with SSH tools, so sleeping that app host stops its loop.
- Unknown after the targeted first-party documentation, release and repository pass: managed_cloud, hybrid, region, expiry, client_disconnect, other_device_continuation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed crates/agent/src/native_agent_server.rs
Finding: The native agent server constructs an agent in the local app process. SSH remoting moves files, tasks and terminals to the user server while model communication stays local; there is no vendor-hosted agent execution mode in this scoped architecture. The local app must remain running for its agent loop.
Workspace isolation: Git worktree: Yes
Workspace isolation: Git worktree
Typed value: Yes.
A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
- Worktree: A thread can select or create a Git worktree to isolate changes from other threads.
- Unknown after the targeted first-party documentation, release and repository pass: vm, process, secrets. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
Workspace isolation: Container: Yes
Workspace isolation: Container
Typed value: Yes.
A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
- Container: Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container.
- Unknown after the targeted first-party documentation, release and repository pass: vm, process, secrets. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Zed builds/opens devcontainers with Docker or Podman and runs tasks, terminals and language servers there. Config changes do not rebuild automatically; users restart and reopen.
Workspace isolation: VM or microVM: No
Workspace isolation: VM or microVM
Typed value: No.
A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
- Vm: No native VM/microVM agent-isolation backend in the inspected sandbox, worktree and devcontainer implementations.
- Unknown after the targeted first-party documentation, release and repository pass: vm, process, secrets. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed crates/agent/src/sandboxing.rs
Finding: The pinned native sandbox supports macOS Seatbelt and Linux/WSL Bubblewrap; worktrees and devcontainers are separate options. The product provides no VM/microVM agent-isolation backend; user servers and Docker host VMs do not add a product VM option.
Parallel runs: Yes cross-checked
Parallel runs
Typed value: Yes.
Multiple agent and terminal threads can run at once, against different projects. Multiple threads can receive independent tasks while previous threads continue. Docs permit multiple concurrent threads but give no numerical product concurrency quota. The Threads Sidebar groups independent conversations by project.
- Concurrent sessions: Multiple agent and terminal threads can run at once, against different projects.
- Unknown after the targeted first-party documentation, release and repository pass: observed_concurrency. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Zed separates its native agent, external ACP agents and terminal threads, with several threads running across projects.
Coordinator: Yes cross-checked
Coordinator
Typed value: Yes.
Native spawn_agent delegates a bounded task to a subagent with a separate context window. agent.subagent_model configures subagent model, thinking, effort and speed; otherwise the native child inherits its parent’s settings. Subagent work returns an outcome to the parent when only its result is needed. The built-in Zed Agent invokes spawn_agent rather than requiring an external coordinator. The parent uses its built-in spawn_agent tool during a task. Resuming a subagent preserves its existing model, reasoning and speed settings.
- Roles: Native spawn_agent delegates a bounded task to a subagent with a separate context window.
- Results return: Subagent work returns an outcome to the parent when only its result is needed.
- Orchestration owner: The built-in Zed Agent invokes spawn_agent rather than requiring an external coordinator.
- Activation: The parent uses its built-in spawn_agent tool during a task.
- Unknown after the targeted first-party documentation, release and repository pass: messages, cross_harness, authentication, live_progress, retained_results, invocation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
Native issue tracking: No
Native issue tracking
Typed value: No.
Zed task records are terminal commands, while agent history consists of conversation threads. The inspected pinned source has no product-owned issue/backlog entity separate from these.
- Task session separation: No product issue/backlog entity; task means terminal command and agent thread means conversation.
- Bounded task/agent/agent_ui schema and dispatcher inspection; external tracker MCP tools are not a native issue system.
-
Finding: Editor tasks are shell commands spawned in terminal tabs. Agent history stores conversation threads, not independent issue/backlog records. A bounded source inspection of task, agent and agent_ui found no native issue entity or issue state machine; GitHub issue operations through MCP remain external.
Unattended triggers: No
Unattended triggers
Typed value: No.
The native agent and agent UI start work from explicit session prompts. Source inspection found no product scheduler or external-event-to-agent run starter.
- Schedules: No scheduled or external-event agent starter in the pinned native agent/UI implementation.
- Events: No scheduled or external-event agent starter in the pinned native agent/UI implementation.
- Recurring tasks: No scheduled or external-event agent starter in the pinned native agent/UI implementation.
- Search covered agent, agent_ui and task modules for cron, schedule and webhook; task hooks are workspace shell setup, not an agent scheduler.
-
Finding: Agent work starts through editor/session prompt actions. The inspected agent and agent_ui modules have no scheduler or webhook-to-agent run service. Editor TaskHook only has CreateWorktree and runs shell setup, not scheduled autonomous agent work.
Shared prompting: No
Shared prompting
Typed value: No.
Collaborators see shared files, cursors, navigation and edits, and guests can view read-only. Persistent channels support shared projects and voice chat. Channels have a shared Markdown notes document for design/status discussion. Multiple editors see each other's cursors and edits in the shared project. Sharing grants project filesystem access; this does not establish simultaneous prompting of the same agent. Shared project code is hosted on the sharer’s machine; unsharing disconnects participants.
- Coprompting: Shared project collaboration does not share a live agent runtime/thread; clients own their own agent connection and stored thread.
- Unknown after the targeted first-party documentation, release and repository pass: coprompting, ordering, conflicts, attribution, plan_restrictions. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed crates/agent/src/native_agent_server.rs
Finding: Each client owns its native agent connection and local thread database. Collaboration shares project edits/cursors, but the pinned collaboration RPC has no shared agent-thread prompt/session service. External agents are disabled for collab guest projects. Shared thread links import copies instead of a live co-prompting session.
Team roles: Yes
Team roles
Typed value: Yes.
Business organizations have Owner, Admin, Billing Manager and Member. Owners/Admins invite by email, change roles and remove non-owner members. Members cannot access billing/admin settings; Billing Managers have billing but no paid Business AI seat. Removal ends organization-managed settings and feature access while personal Zed remains usable. These role rights apply to Zed Business organizations.
- Organization roles: Business organizations have Owner, Admin, Billing Manager and Member.
- Membership: Owners/Admins invite by email, change roles and remove non-owner members.
- Unauthorized access: Members cannot access billing/admin settings; Billing Managers have billing but no paid Business AI seat.
- Revocation: Removal ends organization-managed settings and feature access while personal Zed remains usable.
- Plan scope: These role rights apply to Zed Business organizations.
- Unknown after the targeted first-party documentation, release and repository pass: project_roles, task_roles, sso, provisioning. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business organizations use Owner, Admin, Billing Manager and Member with explicit administration/feature/billing rights. Removing membership revokes org-managed access without disabling personal editor use.
SSO: No
SSO
Typed value: No.
Business organizations have Owner, Admin, Billing Manager and Member. Owners/Admins invite by email, change roles and remove non-owner members. Members cannot access billing/admin settings; Billing Managers have billing but no paid Business AI seat. Removal ends organization-managed settings and feature access while personal Zed remains usable. These role rights apply to Zed Business organizations.
- Sso: The public pricing FAQ explicitly says SSO, SAML and SCIM are planned but unavailable.
- Unknown after the targeted first-party documentation, release and repository pass: project_roles, task_roles, sso, provisioning. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
Pull request workflow: No
Pull request workflow
Typed value: No.
The Git Panel reflects working-tree, staging and branch state, commits and undoes commits, manages branches and linked worktrees, and resolves merge conflicts. Its Create Pull Request action only opens the hosting provider's create page in the browser.
- Pull requests: The Create Pull Request action opens the hosting provider's create-PR page for the tracked branch; Zed does not create or manage the PR itself (v1.27.0 source).
- Pull request creation is handed off to the hosting provider's web page; PR comments, CI response and rebase flows were not established. No installed-product fixture was run.
-
Zed Git Panel create_pull_request
Finding: The Create Pull Request action builds the hosting provider's create-pull-request URL for the tracked upstream branch and opens it in the browser; Zed itself does not create or manage the pull request.
Review interface: Yes cross-checked
Review interface
Typed value: Yes.
Review Changes opens a multi-buffer view of all agent-edited files and changed-line counts. Edited-file accordion and Review Changes navigate the affected files. Users keep or reject individual hunks or the whole agent change set. Review Changes has a documented keyboard shortcut. File/line review remains alongside the agent conversation.
- Change understanding: Review Changes opens a multi-buffer view of all agent-edited files and changed-line counts.
- Evidence navigation: Edited-file accordion and Review Changes navigate the affected files.
- Request changes: Users keep or reject individual hunks or the whole agent change set.
- Keyboard: Review Changes has a documented keyboard shortcut.
- Retained context: File/line review remains alongside the agent conversation.
- Unknown after the targeted first-party documentation, release and repository pass: pending_reviews, comments, shipping_guidance, effort, missed_evidence, narrow_screen, context_switching, roles. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
Product license: Apache-2.0: No. GPL-3.0-or-later cross-checked
Product license: Apache-2.0
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: MIT: No. GPL-3.0-or-later cross-checked
Product license: MIT
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: GPL-3.0: No. GPL-3.0-or-later cross-checked
Product license: GPL-3.0
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: GPL-3.0-or-later: Yes. GPL-3.0-or-later cross-checked
Product license: GPL-3.0-or-later
Typed value: Yes. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: AGPL-3.0: No. GPL-3.0-or-later cross-checked
Product license: AGPL-3.0
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: AGPL-3.0-or-later: No. GPL-3.0-or-later cross-checked
Product license: AGPL-3.0-or-later
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: Elastic License 2.0: No. GPL-3.0-or-later cross-checked
Product license: Elastic License 2.0
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: Custom terms: No. GPL-3.0-or-later cross-checked
Product license: Custom terms
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Product license: Proprietary: No. GPL-3.0-or-later cross-checked
Product license: Proprietary
Typed value: No. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Self-hosting: Yes cross-checked
Self-hosting
Typed value: Yes.
The native editor/agent can use locally hosted llama.cpp, Ollama, LM Studio or customer-compatible endpoints on user infrastructure. Local/native core use can avoid Zed-hosted models; real-time collaboration and Zed-hosted inference remain vendor services. Thread History imports supported external-agent threads; Cursor and Gemini CLI are currently excluded. Editor and own-key AI work without Zed sign-in; hosted models and real-time collaboration require authentication. Archived native threads restore within Zed; a general cross-product transcript export format remains unestablished.
- Self hosting: The native editor/agent can use locally hosted llama.cpp, Ollama, LM Studio or customer-compatible endpoints on user infrastructure.
- Vendor dependencies: Local/native core use can avoid Zed-hosted models; real-time collaboration and Zed-hosted inference remain vendor services.
- Unknown after the targeted first-party documentation, release and repository pass: export, formats, cancellation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Native Zed features work with local llama.cpp, Ollama, LM Studio or an OpenAI-compatible server using configured model/base URL. Endpoint auth is customer controlled.
Telemetry opt-out: Yes cross-checked
Telemetry opt-out
Typed value: Yes.
Zed does not retain prompt or code context by default; opt-in feedback/data sharing is separate. Hosted model requests route through Zed to its providers under no-training contracts. Provider keys saved through Zed are stored in the system keychain, not settings.json. Users inspect client telemetry through zed: open telemetry log. Most hosted inference has ZDR; provider-designated safety-retention models retain prompts/outputs under provider rules, including at least thirty days for named Anthropic Covered Models. Zed-hosted providers commit not to train on prompts/code; optional feedback or edit-prediction training requires explicit sharing. Telemetry documentation names Sentry, Snowflake, Hex and Amplitude for diagnostics/analytics. Zed documents client diagnostics and metrics separately from required hosted-service metadata. telemetry.diagnostics=false and telemetry.metrics=false disable client reports; hosted-service rate-limit/billing metadata is required. Conversation feedback and eligible edit-prediction data are retained only through explicit sharing/opt-in. BYOK, subscription, gateway, local-model, external-agent and terminal paths use their own provider handling.
- Telemetry disclosure: Zed documents client diagnostics and metrics separately from required hosted-service metadata.
- Unknown after the targeted first-party documentation, release and repository pass: artifacts, deletion, residency, telemetry_defaults. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Client metrics and diagnostics can be disabled; hosted AI/collaboration metadata remains necessary for those services.
Free access: Yes cross-checked
Free access
Typed value: Yes.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: Free: Yes. Free cross-checked
Starting price: Free
Typed value: Yes. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: Up to $20/month: No. Free cross-checked
Starting price: Up to $20/month
Typed value: No. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: $21–40/month: No. Free cross-checked
Starting price: $21–40/month
Typed value: No. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: $41–100/month: No. Free cross-checked
Starting price: $41–100/month
Typed value: No. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: Over $100/month: No. Free cross-checked
Starting price: Over $100/month
Typed value: No. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Starting price: Contact for pricing: No. Free cross-checked
Starting price: Contact for pricing
Typed value: No. Free.
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Ongoing free: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route.
- Trial: The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free.
- Card required: The documented trial requires no credit card.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
Pricing model: Free, no paid plan: No
Pricing model: Free, no paid plan
Typed value: No.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Seats: Business costs USD 30 per seat per month; Pro is USD 10 monthly.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
Pricing model: Free plan: Yes
Pricing model: Free plan
Typed value: Yes.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Pricing model: Subscription: Yes
Pricing model: Subscription
Typed value: Yes.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Pricing model: Per seat: Yes
Pricing model: Per seat
Typed value: Yes.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Seats: Business costs USD 30 per seat per month; Pro is USD 10 monthly.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
Pricing model: Usage-based: Yes
Pricing model: Usage-based
Typed value: Yes.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Pricing model: Per machine or compute: No
Pricing model: Per machine or compute
Typed value: No.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Seats: Business costs USD 30 per seat per month; Pro is USD 10 monthly.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
Pricing model: Bring your own key: Yes
Pricing model: Bring your own key
Typed value: Yes.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Pricing model: One-time purchase: No
Pricing model: One-time purchase
Typed value: No.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Seats: Business costs USD 30 per seat per month; Pro is USD 10 monthly.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
Paid plans from: No paid plan: No. $10 or less/month
Paid plans from: No paid plan
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: $10 or less/month: Yes. $10 or less/month
Paid plans from: $10 or less/month
Typed value: Yes. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: $11–20/month: No. $10 or less/month
Paid plans from: $11–20/month
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: $21–40/month: No. $10 or less/month
Paid plans from: $21–40/month
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: $41–100/month: No. $10 or less/month
Paid plans from: $41–100/month
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: Over $100/month: No. $10 or less/month
Paid plans from: Over $100/month
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plans from: Contact sales: No. $10 or less/month
Paid plans from: Contact sales
Typed value: No. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Keeps going with laptop closed: No
Keeps going with laptop closed
Typed value: No.
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
- Managed cloud: No vendor execution mode in the native local-agent/SSH architecture; hosted model inference is separate.
- Client disconnect: The agent/model loop belongs to the controlling local app even with SSH tools, so sleeping that app host stops its loop.
- Unknown after the targeted first-party documentation, release and repository pass: managed_cloud, hybrid, region, expiry, client_disconnect, other_device_continuation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed crates/agent/src/native_agent_server.rs
Finding: The native agent server constructs an agent in the local app process. SSH remoting moves files, tasks and terminals to the user server while model communication stays local; there is no vendor-hosted agent execution mode in this scoped architecture. The local app must remain running for its agent loop.
Open-source license: Yes. GPL-3.0-or-later cross-checked
Open-source license
Typed value: Yes. GPL-3.0-or-later.
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Client license: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked.
- Source visibility: Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked.
- Source revision: The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Any desktop app: Yes
Any desktop app
Typed value: Yes.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Desktop os: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides.
- Browser: The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
Paid plan up to $20/month: Yes. $10 or less/month
Paid plan up to $20/month
Typed value: Yes. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Paid plan up to $40/month: Yes. $10 or less/month
Paid plan up to $40/month
Typed value: Yes. $10 or less/month.
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Currency: Plans and token allowances are quoted in US dollars.
- Included usage: Pro includes five dollars of hosted-model token credit monthly.
- Trial: Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required.
- Model cost: Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
Any phone client: No
Any phone client
Typed value: No.
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Mobile apps: The complete native installation list distributes desktop platforms only; no native iOS/Android build is distributed.
- Mobile browser: The browser version is explicitly not yet available; the current roadmap still places Zed on the Web upcoming.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Zed supported client platforms
Finding: The complete installation list is macOS, Linux and Windows. Other platforms are not yet available, and Web is explicitly listed as unavailable.
All public findings
Findings may summarize a source. They are not presented as verbatim statements.
zed:ACC-01: The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
zed:ACC-01
The pinned README lists macOS, Linux and Windows downloads; minimum OS versions need the platform guides. The pinned Zed README explicitly lists Web among platforms not yet available; current roadmap still lists Zed on the Web as upcoming. The product-owned zed CLI opens local/remote projects, files and diff views. It is an editor launcher/control interface, not a documented headless agent runner. Native threads support prompting, context, cancellation, editable messages, history and reviewing edited files. Agent completion/waiting can generate visual/audio notifications controlled by agent settings.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_apps, mobile_browser, editor_extensions, cross_device_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
-
Finding: The roadmap lists async collaboration in progress and Web and Plan Mode next; undated plans are not current capabilities.
-
Finding: Product-owned zed CLI opens files/directories, remote projects and diff views; wait/reuse/add options integrate with other tools. This is an editor launcher, not a headless agent API.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
-
Zed supported client platforms
Finding: The complete installation list is macOS, Linux and Windows. Other platforms are not yet available, and Web is explicitly listed as unavailable.
zed:ACC-02: Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
zed:ACC-02
Zed Agent runs alongside ACP-integrated External Agents. Documented integrations include Claude Agent/Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Pi, Copilot and Poolside; the current official registry also supplies named Grok Build and Google Antigravity installations. Registry agents appear as managed threads in Agent Panel/sidebar, while each external agent owns its runtime and authentication. Custom ACP executable/arguments/environment configurations remain supported, so this is not a closed list.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed separates its native agent, external ACP agents and terminal threads, with several threads running across projects.
-
Finding: Native Zed AI has hosted, API, subscription, gateway and local model setup routes; these settings do not configure external agents.
-
Finding: Zed ACP registry supports Claude, Codex, Gemini CLI, OpenCode, Cursor and Pi among others. Custom ACP command/args/env adapters are configurable. External auth/tools/retention remain agent owned.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
-
Zed ACP Registry installation and agent threads
Finding: Zed installs registry agents through zed: acp registry or Agent Settings/Add Agent. Installed agents appear in the Agent Panel and Threads Sidebar new-thread menu. Zed manages their threads while each ACP agent owns its runtime and authentication; the common-agent list is explicitly not exhaustive.
-
Grok Build in the Zed ACP Registry
Finding: Zed provides a named Grok Build ACP Registry listing with install/select instructions and the native xAI CLI command npx @xai-official/grok agent stdio. This is the coding-agent runtime, not selecting a Grok model in Zed Agent.
-
Google Antigravity in the Zed ACP Registry
Finding: Zed provides a named Google Antigravity ACP Registry listing with a binary distribution and instructions to install/select it or configure its agy_acp_server.par command as an agent server.
zed:ACC-03: A separate existing-subscription setup route is documented; support must be checked per provider and agent path. Zed Agent supports provider API access; provider-specific setup applies. Zed-hosted models use Zed account billing separately from own provider accounts. Local or self-hosted model endpoints are a documented Zed Agent route. Provider and external-agent authentication are configured separately; native Zed provider settings do not configure external-agent credentials. The editor itself needs no authentication; collaboration features require sign-in. ChatGPT Plus/Pro is a native subscription provider; Claude Pro/Max uses Claude external/CLI paths; OpenCode requires API keys and Cursor uses external/CLI paths. cross-checked
zed:ACC-03
A separate existing-subscription setup route is documented; support must be checked per provider and agent path. Zed Agent supports provider API access; provider-specific setup applies. Zed-hosted models use Zed account billing separately from own provider accounts. Local or self-hosted model endpoints are a documented Zed Agent route. Provider and external-agent authentication are configured separately; native Zed provider settings do not configure external-agent credentials. The editor itself needs no authentication; collaboration features require sign-in. ChatGPT Plus/Pro is a native subscription provider; Claude Pro/Max uses Claude external/CLI paths; OpenCode requires API keys and Cursor uses external/CLI paths.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Native Zed AI has hosted, API, subscription, gateway and local model setup routes; these settings do not configure external agents.
-
Finding: Signed-in users can simultaneously edit a shared project and use channels and voice; project sharing exposes its local filesystem.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
-
Finding: ChatGPT Plus/Pro authenticates as a native Zed provider without API key; Claude Pro/Max uses external Claude or CLI. OpenCode uses API keys; Cursor uses its external/CLI route.
zed:ACC-04: The CLI opens arbitrary files or directories as editor workspaces; no Git remote is required for opening them. The Remote Projects dialog checks SSH connectivity, downloads the remote server and prompts for a project path. cross-checked
zed:ACC-04
The CLI opens arbitrary files or directories as editor workspaces; no Git remote is required for opening them. The Remote Projects dialog checks SSH connectivity, downloads the remote server and prompts for a project path.
- Unknown after the targeted first-party documentation, release and repository pass: first_task_setup, local_git_without_remote, hosted_git, prerequisite_guidance, managed_cloud_setup. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Local UI and model communication remain local while source files, language servers, tasks and terminals run on an SSH server.
-
Finding: Product-owned zed CLI opens files/directories, remote projects and diff views; wait/reuse/add options integrate with other tools. This is an editor launcher, not a headless agent API.
zed:ACC-05: zed [OPTIONS] [PATHS] opens files/directories and supports --wait, --diff, --new, --add and --reuse. CLI opens workspaces, waits for files/windows to close and presents diff views. The launcher is local; editor use does not itself require hosted AI authentication. cross-checked
zed:ACC-05
zed [OPTIONS] [PATHS] opens files/directories and supports --wait, --diff, --new, --add and --reuse. CLI opens workspaces, waits for files/windows to close and presents diff views. The launcher is local; editor use does not itself require hosted AI authentication.
- Unknown after the targeted first-party documentation, release and repository pass: api, sdk, mcp_server, webhooks, stability. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Product-owned zed CLI opens files/directories, remote projects and diff views; wait/reuse/add options integrate with other tools. This is an editor launcher, not a headless agent API.
zed:ACC-06: Zed native agents use locally spawned or remote MCP servers. MCP extensions can expose tool integrations. Configured MCP servers add custom tools beyond native tools. Zed recognizes multiple project formats, including .cursorrules, .clinerules, Copilot instructions, AGENTS.md, CLAUDE.md and GEMINI.md. Remote MCP accepts authorization headers or prompts standard OAuth when no Authorization header is configured. Native MCP tools use agent.tool_permissions and per-tool rules. Users configure servers in Zed; external agents may receive forwarded ACP MCP or use their own config. cross-checked
zed:ACC-06
Zed native agents use locally spawned or remote MCP servers. MCP extensions can expose tool integrations. Configured MCP servers add custom tools beyond native tools. Zed recognizes multiple project formats, including .cursorrules, .clinerules, Copilot instructions, AGENTS.md, CLAUDE.md and GEMINI.md. Remote MCP accepts authorization headers or prompts standard OAuth when no Authorization header is configured. Native MCP tools use agent.tool_permissions and per-tool rules. Users configure servers in Zed; external agents may receive forwarded ACP MCP or use their own config.
- Documentation establishes the named facts; release, plan and execution boundaries remain as stated in the atomic answers. No hands-on fixture was run.
-
Finding: Zed connects local/remote MCP servers through configuration/extensions with headers or OAuth. Native profiles and per-tool permission rules apply; ACP forwarding and external native config are separate.
-
Finding: Personal AGENTS.md and compatible project instruction files provide persistent context; project rules override personal instructions. External/terminal agents have their own loaders.
zed:ACC-07: Zed uses Rust and the GPU-accelerated GPUI framework for its editor/client. Vendor developer docs identify gpui, editor, project, workspace and cli crates; no installed-build performance test was run. cross-checked
zed:ACC-07
Zed uses Rust and the GPU-accelerated GPUI framework for its editor/client. Vendor developer docs identify gpui, editor, project, workspace and cli crates; no installed-build performance test was run.
- Unknown after the targeted first-party documentation, release and repository pass: mobile_implementation, hybrids. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Vendor developer documentation identifies Rust crates, GPU-accelerated GPUI, native editor/project/workspace components and a separate collaboration server.
zed:PLAN-01: Users create native or installed external-agent threads from the panel/sidebar. The Agent Panel accepts project-context prompts. Sent messages can be edited and resubmitted with revised prompts/context. cross-checked
zed:PLAN-01
Users create native or installed external-agent threads from the panel/sidebar. The Agent Panel accepts project-context prompts. Sent messages can be edited and resubmitted with revised prompts/context.
- Unknown after the targeted first-party documentation, release and repository pass: external_import, chat_git_triggers, authoritative_tracker, stable_ids, sync, external_roundtrips. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:PLAN-04: Zed Agent loads personal AGENTS.md and the first matching compatible project instruction file. Native SKILL.md packages load on demand or through /slash and @ mentions. Project guidance can use .rules, AGENTS.md, CLAUDE.md, GEMINI.md and other documented compatibility paths. Skills are global under ~/.agents/skills or project-local .agents/skills. Imported/shared skills show their body and target path; a recipient must explicitly save a shared skill. Users edit the personal/project instruction files to change persistent guidance. Skills Manager deletes the installed folder. cross-checked
zed:PLAN-04
Zed Agent loads personal AGENTS.md and the first matching compatible project instruction file. Native SKILL.md packages load on demand or through /slash and @ mentions. Project guidance can use .rules, AGENTS.md, CLAUDE.md, GEMINI.md and other documented compatibility paths. Skills are global under ~/.agents/skills or project-local .agents/skills. Imported/shared skills show their body and target path; a recipient must explicitly save a shared skill. Users edit the personal/project instruction files to change persistent guidance. Skills Manager deletes the installed folder.
- Unknown after the targeted first-party documentation, release and repository pass: memory, freshness, takeover_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Personal AGENTS.md and compatible project instruction files provide persistent context; project rules override personal instructions. External/terminal agents have their own loaders.
-
Finding: SKILL.md packages are created/imported, scoped global/project, editable/deletable and shared as self-contained links requiring explicit recipient save. Model-loaded skills follow tool permissions; manual invocation is explicit.
zed:PLAN-05: Native find_path and regex grep navigate project files. Files, directories, symbols, previous threads, diagnostics, diffs and fetched URLs can be attached as context. Vision-capable providers accept images via mention, drag/drop or paste. search_web is Pro-only with the Zed provider; other configurations can use a search MCP server. Pasted code selections retain their file-context mention. cross-checked
zed:PLAN-05
Native find_path and regex grep navigate project files. Files, directories, symbols, previous threads, diagnostics, diffs and fetched URLs can be attached as context. Vision-capable providers accept images via mention, drag/drop or paste. search_web is Pro-only with the Zed provider; other configurations can use a search MCP server. Pasted code selections retain their file-context mention.
- Unknown after the targeted first-party documentation, release and repository pass: indexing, freshness, api_retrieval, conflict_detection, review_enforcement. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
zed:PLAN-06: Zed task records are terminal commands, while agent history consists of conversation threads. The inspected pinned source has no product-owned issue/backlog entity separate from these.
zed:PLAN-06
Zed task records are terminal commands, while agent history consists of conversation threads. The inspected pinned source has no product-owned issue/backlog entity separate from these.
- Bounded task/agent/agent_ui schema and dispatcher inspection; external tracker MCP tools are not a native issue system.
-
Finding: Editor tasks are shell commands spawned in terminal tabs. Agent history stores conversation threads, not independent issue/backlog records. A bounded source inspection of task, agent and agent_ui found no native issue entity or issue state machine; GitHub issue operations through MCP remain external.
zed:PLAN-08: Persistent channel Markdown notes can hold design/status ideas, but are not a typed requirements object. Channel members collaboratively edit Markdown notes. Channel notes can be opened without joining its call. cross-checked
zed:PLAN-08
Persistent channel Markdown notes can hold design/status ideas, but are not a typed requirements object. Channel members collaboratively edit Markdown notes. Channel notes can be opened without joining its call.
- Unknown after the targeted first-party documentation, release and repository pass: agent_edits, revisions, approvals, implementation_links, drift_detection, review_integration. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Channels offer real-time shared code editing, voice and Markdown notes. Guests are read-only unless temporarily granted write access. Unsharing disconnects collaborators; terminal following is not supported.
zed:COORD-01: Multiple agent and terminal threads can run at once, against different projects. Multiple threads can receive independent tasks while previous threads continue. Docs permit multiple concurrent threads but give no numerical product concurrency quota. The Threads Sidebar groups independent conversations by project. cross-checked
zed:COORD-01
Multiple agent and terminal threads can run at once, against different projects. Multiple threads can receive independent tasks while previous threads continue. Docs permit multiple concurrent threads but give no numerical product concurrency quota. The Threads Sidebar groups independent conversations by project.
- Unknown after the targeted first-party documentation, release and repository pass: observed_concurrency. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Zed separates its native agent, external ACP agents and terminal threads, with several threads running across projects.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
zed:COORD-03: Native spawn_agent delegates a bounded task to a subagent with a separate context window. agent.subagent_model configures subagent model, thinking, effort and speed; otherwise the native child inherits its parent’s settings. Subagent work returns an outcome to the parent when only its result is needed. The built-in Zed Agent invokes spawn_agent rather than requiring an external coordinator. The parent uses its built-in spawn_agent tool during a task. Resuming a subagent preserves its existing model, reasoning and speed settings. cross-checked
zed:COORD-03
Native spawn_agent delegates a bounded task to a subagent with a separate context window. agent.subagent_model configures subagent model, thinking, effort and speed; otherwise the native child inherits its parent’s settings. Subagent work returns an outcome to the parent when only its result is needed. The built-in Zed Agent invokes spawn_agent rather than requiring an external coordinator. The parent uses its built-in spawn_agent tool during a task. Resuming a subagent preserves its existing model, reasoning and speed settings.
- Unknown after the targeted first-party documentation, release and repository pass: messages, cross_harness, authentication, live_progress, retained_results, invocation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
-
Finding: Subagents inherit model/effort/speed by default or use agent.subagent_model. Requested models preserve configured controls with documented fallback; compaction defaults to ninety percent context.
zed:COORD-04: The native agent and agent UI start work from explicit session prompts. Source inspection found no product scheduler or external-event-to-agent run starter.
zed:COORD-04
The native agent and agent UI start work from explicit session prompts. Source inspection found no product scheduler or external-event-to-agent run starter.
- Search covered agent, agent_ui and task modules for cron, schedule and webhook; task hooks are workspace shell setup, not an agent scheduler.
-
Finding: Agent work starts through editor/session prompt actions. The inspected agent and agent_ui modules have no scheduler or webhook-to-agent run service. Editor TaskHook only has CreateWorktree and runs shell setup, not scheduled autonomous agent work.
zed:COORD-05: A stop button interrupts execution; a distinct durable pause mechanism is not established. The panel documents an immediate stop action, separate from queued follow-up messages. Queued messages can interrupt Zed Agent at its next step using Steer; external-agent turn boundaries are not detected. Waiting/completion notifications can be enabled visually and audibly. Permission-gated tool actions request human confirmation before running. Native threads retain conversation history and can start fresh from a summary. cross-checked
zed:COORD-05
A stop button interrupts execution; a distinct durable pause mechanism is not established. The panel documents an immediate stop action, separate from queued follow-up messages. Queued messages can interrupt Zed Agent at its next step using Steer; external-agent turn boundaries are not detected. Waiting/completion notifications can be enabled visually and audibly. Permission-gated tool actions request human confirmation before running. Native threads retain conversation history and can start fresh from a summary.
- Unknown after the targeted first-party documentation, release and repository pass: resume, escalation, visible_errors, chat_terminal_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:COORD-07: Native Write, Ask and Minimal profiles provide different tool sets. Native thread settings and subagent model/effort/speed can be configured separately. Later parent model/effort changes do not update already configured/requested child models. The panel’s profile selector exposes Configure and custom-profile creation. cross-checked
zed:COORD-07
Native Write, Ask and Minimal profiles provide different tool sets. Native thread settings and subagent model/effort/speed can be configured separately. Later parent model/effort changes do not update already configured/requested child models. The panel’s profile selector exposes Configure and custom-profile creation.
- Unknown after the targeted first-party documentation, release and repository pass: quick_task, growing_epic, natural_language, live_roles, retained_relationships, auth_friction. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Subagents inherit model/effort/speed by default or use agent.subagent_model. Requested models preserve configured controls with documented fallback; compaction defaults to ninety percent context.
-
Finding: Write enables edits/commands, Ask is read-only and Minimal has no project tools. Profiles set tool availability and default model; permission rules are separate and external agents do not inherit these profiles.
zed:RUN-01: Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
zed:RUN-01
Native agent terminal tools run shell commands in the project environment. SSH remoting runs code, tools and terminals remotely; model communication remains on the local client, so unattended execution after closing it is not established. Zed Remote Development uses a local UI and customer-hosted headless server connected over SSH. The local client builds/launches the configured devcontainer using installed Docker/Podman. The user supplies and pays for the remote server or devcontainer host. The user supplies local UI hardware and remote host compute.
- Unknown after the targeted first-party documentation, release and repository pass: managed_cloud, hybrid, region, expiry, client_disconnect, other_device_continuation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Local UI and model communication remain local while source files, language servers, tasks and terminals run on an SSH server.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
-
Finding: Zed builds/opens devcontainers with Docker or Podman and runs tasks, terminals and language servers there. Config changes do not rebuild automatically; users restart and reopen.
-
Zed crates/agent/src/native_agent_server.rs
Finding: The native agent server constructs an agent in the local app process. SSH remoting moves files, tasks and terminals to the user server while model communication stays local; there is no vendor-hosted agent execution mode in this scoped architecture. The local app must remain running for its agent loop.
zed:RUN-02: A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
zed:RUN-02
A new agent worktree has its own checkout. New worktrees start detached HEAD; a branch is selected/created separately. A thread can select or create a Git worktree to isolate changes from other threads. Zed opens projects inside devcontainers using Docker or Podman; tasks/terminals run in that container. Zed Agent terminal and fetch tools use OS-enforced resource restrictions; external agents, terminal threads and ordinary tasks are excluded. When enabled, the sandbox permits only configured write locations by default while allowing broad file reads. Enabled terminal sandbox blocks outbound networking until approved; WSL grants are all-or-nothing. Separate worktrees prevent overlapping working files; model conversations remain independent. Native terminal sandbox has OS-level restrictions; Windows requires WSL, and separate ACP/MCP/fetch tools have different boundaries.
- Unknown after the targeted first-party documentation, release and repository pass: vm, process, secrets. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: OS sandboxing covers Zed Agent terminal/fetch tools, not external agents or terminal threads; platform-specific limits apply.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
-
Finding: Zed builds/opens devcontainers with Docker or Podman and runs tasks, terminals and language servers there. Config changes do not rebuild automatically; users restart and reopen.
-
Finding: Native sandbox blocks terminal networking and out-of-scope writes by default when enabled. macOS uses Seatbelt, Linux Bubblewrap and Windows WSL Bubblewrap only. Non-terminal fetch does not inherit terminal network limits.
-
Zed crates/agent/src/sandboxing.rs
Finding: The pinned native sandbox supports macOS Seatbelt and Linux/WSL Bubblewrap; worktrees and devcontainers are separate options. The product provides no VM/microVM agent-isolation backend; user servers and Docker host VMs do not add a product VM option.
zed:RUN-03: Thread History can restore Zed threads; external-agent restoration depends on its integration. Archived native threads can be restored, including their removed linked worktree. Archiving an unused linked worktree saves its Git state; restoration recreates it. External thread import/history support varies; Cursor/Gemini CLI imports are explicitly not supported. cross-checked
zed:RUN-03
Thread History can restore Zed threads; external-agent restoration depends on its integration. Archived native threads can be restored, including their removed linked worktree. Archiving an unused linked worktree saves its Git state; restoration recreates it. External thread import/history support varies; Cursor/Gemini CLI imports are explicitly not supported.
- Unknown after the targeted first-party documentation, release and repository pass: client_disconnect, worker_crash, service_restart, expiry, process_reconnect, context_loss, duplicate_actions, multi_day_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
zed:RUN-04: Project devcontainer.json defines the container image, dependencies and environment. Language servers, tasks and terminals run on the user’s remote server. SSH port forwarding is configurable per remote project. Open in Container builds the image if needed, launches the container and reconnects the project. Changes to devcontainer.json do not automatically rebuild/reload the container; users must stop it and reopen. cross-checked
zed:RUN-04
Project devcontainer.json defines the container image, dependencies and environment. Language servers, tasks and terminals run on the user’s remote server. SSH port forwarding is configurable per remote project. Open in Container builds the image if needed, launches the container and reconnects the project. Changes to devcontainer.json do not automatically rebuild/reload the container; users must stop it and reopen.
- Unknown after the targeted first-party documentation, release and repository pass: caches, snapshots, secrets, duplicate_installs, collisions. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Local UI and model communication remain local while source files, language servers, tasks and terminals run on an SSH server.
-
Finding: Zed builds/opens devcontainers with Docker or Podman and runs tasks, terminals and language servers there. Config changes do not rebuild automatically; users restart and reopen.
zed:RUN-05: Automatic compaction defaults at ninety percent of context; a configurable threshold or /compact is supported. Pro/Business set pre-tax monthly hosted-token spend limits; included credit is separate. Hosted-model use stops when the configured limit is reached until reset or administrator increase. cross-checked
zed:RUN-05
Automatic compaction defaults at ninety percent of context; a configurable threshold or /compact is supported. Pro/Business set pre-tax monthly hosted-token spend limits; included credit is separate. Hosted-model use stops when the configured limit is reached until reset or administrator increase.
- Unknown after the targeted first-party documentation, release and repository pass: time_limits, turn_limits, timeouts, retries, provider_failures. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
-
Finding: Subagents inherit model/effort/speed by default or use agent.subagent_model. Requested models preserve configured controls with documented fallback; compaction defaults to ninety percent context.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
zed:RUN-06: The thread transcript shows streamed responses and tools in use. Panel surfaces active-thread token usage and edited-file/line counts. cross-checked
zed:RUN-06
The thread transcript shows streamed responses and tools in use. Panel surfaces active-thread token usage and edited-file/line counts.
- Unknown after the targeted first-party documentation, release and repository pass: traces, failure_reasons, usage_export, agent_stats, project_stats, time_stats, failure_status, approval_status, reconnect_consistency, performance_method, responsiveness. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:RUN-07: Active native threads show token usage near the model/profile selector. The current thread exposes token counts; complete cross-harness invoice reconciliation is not established. cross-checked
zed:RUN-07
Active native threads show token usage near the model/profile selector. The current thread exposes token counts; complete cross-harness invoice reconciliation is not established.
- Unknown after the targeted first-party documentation, release and repository pass: task_cost, session_cost, child_cost, epic_cost, harness_aggregation, billed_vs_estimated, subscription_allowance, provider_reconciliation, compute_cost, refresh_delay, alerts, export. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:TEAM-01: Collaborators see shared files, cursors, navigation and edits, and guests can view read-only. Persistent channels support shared projects and voice chat. Channels have a shared Markdown notes document for design/status discussion. Multiple editors see each other's cursors and edits in the shared project. Sharing grants project filesystem access; this does not establish simultaneous prompting of the same agent. Shared project code is hosted on the sharer’s machine; unsharing disconnects participants.
zed:TEAM-01
Collaborators see shared files, cursors, navigation and edits, and guests can view read-only. Persistent channels support shared projects and voice chat. Channels have a shared Markdown notes document for design/status discussion. Multiple editors see each other's cursors and edits in the shared project. Sharing grants project filesystem access; this does not establish simultaneous prompting of the same agent. Shared project code is hosted on the sharer’s machine; unsharing disconnects participants.
- Unknown after the targeted first-party documentation, release and repository pass: coprompting, ordering, conflicts, attribution, plan_restrictions. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Signed-in users can simultaneously edit a shared project and use channels and voice; project sharing exposes its local filesystem.
-
Finding: Channels offer real-time shared code editing, voice and Markdown notes. Guests are read-only unless temporarily granted write access. Unsharing disconnects collaborators; terminal following is not supported.
-
Zed crates/agent/src/native_agent_server.rs
Finding: Each client owns its native agent connection and local thread database. Collaboration shares project edits/cursors, but the pinned collaboration RPC has no shared agent-thread prompt/session service. External agents are disabled for collab guest projects. Shared thread links import copies instead of a live co-prompting session.
zed:TEAM-02: Business organizations have Owner, Admin, Billing Manager and Member. Owners/Admins invite by email, change roles and remove non-owner members. Members cannot access billing/admin settings; Billing Managers have billing but no paid Business AI seat. Removal ends organization-managed settings and feature access while personal Zed remains usable. These role rights apply to Zed Business organizations.
zed:TEAM-02
Business organizations have Owner, Admin, Billing Manager and Member. Owners/Admins invite by email, change roles and remove non-owner members. Members cannot access billing/admin settings; Billing Managers have billing but no paid Business AI seat. Removal ends organization-managed settings and feature access while personal Zed remains usable. These role rights apply to Zed Business organizations.
- Unknown after the targeted first-party documentation, release and repository pass: project_roles, task_roles, sso, provisioning. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Business organizations use Owner, Admin, Billing Manager and Member with explicit administration/feature/billing rights. Removing membership revokes org-managed access without disabling personal editor use.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
zed:TEAM-03: Native terminal commands are matched against regex allow/deny/confirm rules, including chained subcommands. Profiles determine tool availability; permission rules separately govern confirmations. Enabled native terminal sandbox blocks network by default and requests host-specific or unrestricted access. Sandbox limits writes/network, but terminal reads can include much of the filesystem; it is not a claim that every credential is unreadable. Rules persist in the user/project settings file until changed. always_allow or a global allow default auto-approves eligible actions. always_deny blocks matching actions even when global permission default allows. macOS/Linux host-specific access uses an HTTP proxy; SSH/raw-socket tools may not work under those host restrictions. The documented global tool-permissions default is confirm. Ask removes edit/terminal tools for read-only questions; Minimal uses no project tools. Hardcoded security rules and always_deny outrank allow; always_confirm can force approval. cross-checked
zed:TEAM-03
Native terminal commands are matched against regex allow/deny/confirm rules, including chained subcommands. Profiles determine tool availability; permission rules separately govern confirmations. Enabled native terminal sandbox blocks network by default and requests host-specific or unrestricted access. Sandbox limits writes/network, but terminal reads can include much of the filesystem; it is not a claim that every credential is unreadable. Rules persist in the user/project settings file until changed. always_allow or a global allow default auto-approves eligible actions. always_deny blocks matching actions even when global permission default allows. macOS/Linux host-specific access uses an HTTP proxy; SSH/raw-socket tools may not work under those host restrictions. The documented global tool-permissions default is confirm. Ask removes edit/terminal tools for read-only questions; Minimal uses no project tools. Hardcoded security rules and always_deny outrank allow; always_confirm can force approval.
- Unknown after the targeted first-party documentation, release and repository pass: delivery, approver_identity, revoke. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: OS sandboxing covers Zed Agent terminal/fetch tools, not external agents or terminal threads; platform-specific limits apply.
-
Finding: Tool rules support allow, deny and confirm; deny takes precedence and built-in security rules cannot be overridden.
-
Finding: Write enables edits/commands, Ask is read-only and Minimal has no project tools. Profiles set tool availability and default model; permission rules are separate and external agents do not inherit these profiles.
-
Finding: Native tools support allow/deny/confirm patterns. Hardcoded protections and always_deny outrank auto-approve; chained shell commands are individually checked. ACP/terminal-agent boundaries remain separate.
-
Finding: Native sandbox blocks terminal networking and out-of-scope writes by default when enabled. macOS uses Seatbelt, Linux Bubblewrap and Windows WSL Bubblewrap only. Non-terminal fetch does not inherit terminal network limits.
zed:TEAM-04: Thread History retains archived conversations across projects until deleted. Permanent deletion removes conversation history and associated worktree data, and cannot be undone. cross-checked
zed:TEAM-04
Thread History retains archived conversations across projects until deleted. Permanent deletion removes conversation history and associated worktree data, and cannot be undone.
- Unknown after the targeted first-party documentation, release and repository pass: actions, approvals, edits, discussion, decisions, cross_user_persistence, export, mutability, integrity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Independent threads have history and optional worktrees. Steer works for Zed Agent, while external-agent history and token support vary.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
zed:TEAM-05: Review Changes opens a multi-buffer view of all agent-edited files and changed-line counts. Edited-file accordion and Review Changes navigate the affected files. Users keep or reject individual hunks or the whole agent change set. Review Changes has a documented keyboard shortcut. File/line review remains alongside the agent conversation. cross-checked
zed:TEAM-05
Review Changes opens a multi-buffer view of all agent-edited files and changed-line counts. Edited-file accordion and Review Changes navigate the affected files. Users keep or reject individual hunks or the whole agent change set. Review Changes has a documented keyboard shortcut. File/line review remains alongside the agent conversation.
- Unknown after the targeted first-party documentation, release and repository pass: pending_reviews, comments, shipping_guidance, effort, missed_evidence, narrow_screen, context_switching, roles. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:SHIP-01: Project Diff shows repository changes. cross-checked
zed:SHIP-01
Project Diff shows repository changes.
- Retained artifacts and outputs: first-party sources in this pass did not establish documents, screenshots, recordings, previews, test_outputs, reviewer_outputs, gallery, task_links, revision_links, persistence, child_findings, prompt_leakage. No draft PR, failed-gate, stale-approval, rollback or six-task batch-review fixture was run. Described actions do not prove delivery authority.
-
Finding: Zed displays working tree, staging and branch state with native Git controls; unsupported operations can use the terminal.
zed:SHIP-02: The native terminal tool can run repository tests and return command output. The diagnostics tool reads file or project errors/warnings after edits. Tool access permits tests/diagnostics; no universal mandatory delivery check is established. cross-checked
zed:SHIP-02
The native terminal tool can run repository tests and return command output. The diagnostics tool reads file or project errors/warnings after edits. Tool access permits tests/diagnostics; no universal mandatory delivery check is established.
- Unknown after the targeted first-party documentation, release and repository pass: lint, ci, independent_review, spec_checks, failure_blocking, requirement_conflicts, configured_policy, enforced_policy. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Native tools include grep/find/read, diagnostics, fetch/search_web, file edits, terminal and spawn_agent with a separate context window. Pro/Zed-provider-only web search is distinct from MCP search.
zed:SHIP-03: The Git Panel reflects working-tree, staging and branch state, commits and undoes commits, manages branches and linked worktrees, and resolves merge conflicts. Its Create Pull Request action only opens the hosting provider's create page in the browser. cross-checked
zed:SHIP-03
The Git Panel reflects working-tree, staging and branch state, commits and undoes commits, manages branches and linked worktrees, and resolves merge conflicts. Its Create Pull Request action only opens the hosting provider's create page in the browser.
- Pull request creation is handed off to the hosting provider's web page; PR comments, CI response and rebase flows were not established. No installed-product fixture was run.
-
Finding: Zed displays working tree, staging and branch state with native Git controls; unsupported operations can use the terminal.
-
Zed Git documentation (v1.27.0)
Finding: The Git Panel commits staged changes and can undo the last commit, creates, switches and deletes branches, creates and switches linked worktrees from a picker, and shows merge conflicts with per-conflict resolution buttons.
-
Zed Git Panel create_pull_request
Finding: The Create Pull Request action builds the hosting provider's create-pull-request URL for the tracked upstream branch and opens it in the browser; Zed itself does not create or manage the pull request.
zed:SHIP-06: Multi-buffer review groups the thread’s changed files. The review surface shows actual file diffs. Review controls keep or reject hunks or all changes. cross-checked
zed:SHIP-06
Multi-buffer review groups the thread’s changed files. The review surface shows actual file diffs. Review controls keep or reject hunks or all changes.
- Unknown after the targeted first-party documentation, release and repository pass: review_queue, summaries, risk_size_filters, findings, request_changes, exact_revision_approval, dependency_queues, stale_approval_invalidation, check_deduplication. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Panel supports queued/steer messages, native thread history and summaries, per-hunk keep/reject, multi-buffer review, context attachments and token counts. External integration feature parity varies.
zed:OWN-01: At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components. cross-checked
zed:OWN-01
At commit 0dea8c63d9a5d747b3d0903adb110a9def1d1020, Zed's README states primarily GPL-3.0-or-later with Apache-2.0 components where marked. Zed’s public repository and build guides cover macOS, Linux and Windows. Main source is GPL-3.0-or-later with Apache-2.0 components where marked. The license statement is scoped to the pinned README revision, not an assertion about all hosted components.
- Unknown after the targeted first-party documentation, release and repository pass: server_license, hosted_extensions, hosted_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: This pinned README states primarily GPL-3.0-or-later with Apache-2.0 components where marked; hosted-service parity does not follow.
zed:OWN-02: The native editor/agent can use locally hosted llama.cpp, Ollama, LM Studio or customer-compatible endpoints on user infrastructure. Local/native core use can avoid Zed-hosted models; real-time collaboration and Zed-hosted inference remain vendor services. Thread History imports supported external-agent threads; Cursor and Gemini CLI are currently excluded. Editor and own-key AI work without Zed sign-in; hosted models and real-time collaboration require authentication. Archived native threads restore within Zed; a general cross-product transcript export format remains unestablished. cross-checked
zed:OWN-02
The native editor/agent can use locally hosted llama.cpp, Ollama, LM Studio or customer-compatible endpoints on user infrastructure. Local/native core use can avoid Zed-hosted models; real-time collaboration and Zed-hosted inference remain vendor services. Thread History imports supported external-agent threads; Cursor and Gemini CLI are currently excluded. Editor and own-key AI work without Zed sign-in; hosted models and real-time collaboration require authentication. Archived native threads restore within Zed; a general cross-product transcript export format remains unestablished.
- Unknown after the targeted first-party documentation, release and repository pass: export, formats, cancellation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Threads run independently with different native/external agents, own history and context. History archives/restores/deletes, agent-dependent imports and detached-HEAD worktree save/restore are documented.
-
Finding: Native Zed features work with local llama.cpp, Ollama, LM Studio or an OpenAI-compatible server using configured model/base URL. Endpoint auth is customer controlled.
-
Finding: Editor use needs no login. Own-key AI can work without Zed sign-in; hosted models and real-time collaboration require Zed’s GitHub OAuth account.
zed:OWN-03: Zed does not retain prompt or code context by default; opt-in feedback/data sharing is separate. Hosted model requests route through Zed to its providers under no-training contracts. Provider keys saved through Zed are stored in the system keychain, not settings.json. Users inspect client telemetry through zed: open telemetry log. Most hosted inference has ZDR; provider-designated safety-retention models retain prompts/outputs under provider rules, including at least thirty days for named Anthropic Covered Models. Zed-hosted providers commit not to train on prompts/code; optional feedback or edit-prediction training requires explicit sharing. Telemetry documentation names Sentry, Snowflake, Hex and Amplitude for diagnostics/analytics. Zed documents client diagnostics and metrics separately from required hosted-service metadata. telemetry.diagnostics=false and telemetry.metrics=false disable client reports; hosted-service rate-limit/billing metadata is required. Conversation feedback and eligible edit-prediction data are retained only through explicit sharing/opt-in. BYOK, subscription, gateway, local-model, external-agent and terminal paths use their own provider handling. cross-checked
zed:OWN-03
Zed does not retain prompt or code context by default; opt-in feedback/data sharing is separate. Hosted model requests route through Zed to its providers under no-training contracts. Provider keys saved through Zed are stored in the system keychain, not settings.json. Users inspect client telemetry through zed: open telemetry log. Most hosted inference has ZDR; provider-designated safety-retention models retain prompts/outputs under provider rules, including at least thirty days for named Anthropic Covered Models. Zed-hosted providers commit not to train on prompts/code; optional feedback or edit-prediction training requires explicit sharing. Telemetry documentation names Sentry, Snowflake, Hex and Amplitude for diagnostics/analytics. Zed documents client diagnostics and metrics separately from required hosted-service metadata. telemetry.diagnostics=false and telemetry.metrics=false disable client reports; hosted-service rate-limit/billing metadata is required. Conversation feedback and eligible edit-prediction data are retained only through explicit sharing/opt-in. BYOK, subscription, gateway, local-model, external-agent and terminal paths use their own provider handling.
- Unknown after the targeted first-party documentation, release and repository pass: artifacts, deletion, residency, telemetry_defaults. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Zed describes no default retention of prompts/code and separates hosted commitments from own-provider terms; some models have safety-retention exceptions.
-
Finding: Client metrics and diagnostics can be disabled; hosted AI/collaboration metadata remains necessary for those services.
-
Finding: Zed normally retains no prompt/code context; provider contracts prohibit training but designated safety-retention models are exceptions to ZDR. Feedback/training sharing is explicit opt-in. BYOK/external/terminal paths follow their own providers.
-
Finding: Client metrics/crashes can be disabled independently. Hosted-service metadata needed for billing/rate limiting remains required; account-linked random IDs and inspectable logs are documented.
zed:OWN-04: Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
zed:OWN-04
Plans and token allowances are quoted in US dollars. Pro's spend-limit example states a $10/month subscription with $5 included token credit, checked 7 November 2026. Pro includes five dollars of hosted-model token credit monthly. Fourteen-day Pro trial includes five dollars of designated-model usage and predictions; no card required. Hosted tokens use provider list price plus 10%; Pro's default incremental spend limit is $10 pre-tax and can be set to zero. Hosted inference is provider list pricing plus ten percent.
- Unknown after the targeted first-party documentation, release and repository pass: seats, seat_minimum, workspace_fees, credits, taxes, compute_cost. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
-
Finding: Business costs USD 30 per seat per month. Personal is free and Pro is USD 10 monthly. The complete plan list has no compute or one-time license fee. The FAQ explicitly says SSO, SAML and SCIM are planned, not currently available.
zed:OWN-06: Zed publishes a stable editor release channel alongside other channels. The official stable-release page contains current release notes; this was read on 2026-11-07. cross-checked
zed:OWN-06
Zed publishes a stable editor release channel alongside other channels. The official stable-release page contains current release notes; this was read on 2026-11-07.
- Unknown after the targeted first-party documentation, release and repository pass: access_restrictions, support, response_commitments, status_history, deprecation, export_options. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: Stable-release notes publish current shipped editor fixes and new features. This source pass did not install or benchmark the release.
zed:OWN-08: Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro. cross-checked
zed:OWN-08
Zed Free supports the editor, own API access and external agents; Zed-hosted models require a paid or trial route. The fourteen-day trial includes $5 GPT-6 Luna usage and unlimited edit predictions, then reverts to Free. Free editor supports user API keys and external agents; paid hosted-model costs remain separate. The documented trial requires no credit card. Pro monthly token allowance is five dollars; the student plan is capped at ten dollars monthly. Trial automatically reverts to Free after fourteen days; it does not auto-enroll in paid Pro.
- Unknown after the targeted first-party documentation, release and repository pass: credits_trial, self_host_cost, paid_only, early_access. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.
zed:OWN-09: Hosted-model usage is metered in tokens and converted to dollars. Included hosted-model credit resets on the monthly billing date. Usage above the allowance is billed at hosted-model rates, bounded by configured spend limits. Hosted usage is token based at provider list price plus 10%. Public base price, included token allowance, model markup and spend-limit mechanics are documented; fixed workload costs remain unknown. cross-checked
zed:OWN-09
Hosted-model usage is metered in tokens and converted to dollars. Included hosted-model credit resets on the monthly billing date. Usage above the allowance is billed at hosted-model rates, bounded by configured spend limits. Hosted usage is token based at provider list price plus 10%. Public base price, included token allowance, model markup and spend-limit mechanics are documented; fixed workload costs remain unknown.
- Unknown after the targeted first-party documentation, release and repository pass: compute_markup, minimum_commitment, cancellation, sales_only_unknowns. No configured product fixture was run; documentation does not establish successful hands-on behavior.
-
Finding: The editor works without an account. Pro includes $5 monthly usage, metered at provider list price plus 10%, with configurable spend limits.
-
Finding: Free editor includes own-key and external-agent use. Pro is ten dollars monthly with five token dollars; hosted usage is provider list plus ten percent. Trial is fourteen days/no card and shares balance with Delta; no Delta execution capability is inferred.