Keep Zed.
Give it your machines.
AI Commander is not an alternative to Zed, Cursor, Claude Code, Codex, ChatGPT, Gemini, Grok, Windsurf, Continue, Cline, or Roo Code. It is the remote-machine harness those tools plug into over MCP: outbound-only shell, detached jobs that outlive the chat, and file transfer on hardware you already own.
- Install the host agent on the machine you want your AI to control. It shows a private machine code.
- Connect your AI: paste this prompt, then replace
AIC-…with your code —Help me connect this AI client to AI Commander using https://aicommander.dev/start.md. Then check the status of my machine: AIC-…— or add the connector URLhttps://aicommander.dev/mcpand sign in with an email code (per-client steps).
Sibling harness pages: Cursor · Aider · Cline · Roo Code · Zed · OpenCode · Qwen Code · Continue · Windsurf · Claude Code · Grok · Codex · ChatGPT · Gemini.
Connect Zed (context_servers)
Zed is a high-performance code editor with a built-in Agent Panel and native MCP support
(Zed calls them context servers).
Add AI Commander under context_servers in Zed’s settings
(Settings → AI → MCP Servers, or command palette zed: open settings file):
{
"context_servers": {
"aicommander": {
"command": "npx",
"args": ["-y", "@aicommander/mcp"],
"env": {
"AICOMMANDER_SERVER": "https://aicommander.dev"
}
}
}
}
Prefer a remote HTTP MCP endpoint? Zed also accepts a URL entry (OAuth / DCR on first connect):
{
"context_servers": {
"aicommander": {
"url": "https://aicommander.dev/mcp"
}
}
}
Full client matrix (Zed, Cursor, Continue, Windsurf, Cline, Roo Code, Claude Code, Codex, ChatGPT, Gemini, Grok, REST): Connect your AI client.
Over the remote URL, first connect may open a one-time browser sign-in (OAuth Dynamic Client Registration).
The npx bridge does not sign in through a browser: without AICOMMANDER_TOKEN in its env it connects anonymously with a code-only first hour; set it to your API key to bind it to your account.
On the remote URL, append ?anonymous=1 to skip the sign-in, for a code-only first hour, where your client supports it.
What you get that the chat alone cannot
- Remote shell on Linux, macOS, and Windows — by session code or saved alias.
- Detached jobs for builds, training, and headless agents on the box that has the repo.
- Fleet list + GPU facts so Zed picks the right machine before it burns an hour.
- File courier (Pro) for checkpoints and diffs up to 100 MiB.
Not a replacement — a multiplier
| You already use | AI Commander adds |
|---|---|
| Zed | context_servers in settings.json — local npx @aicommander/mcp or remote https://aicommander.dev/mcp |
| Cursor | npx @aicommander/mcp stdio bridge → remote exec + jobs |
| Continue | ~/.continue/config.yaml / .continue/mcpServers/ + stdio bridge |
| Cline / Roo Code | MCP settings JSON + npx @aicommander/mcp |
| Windsurf (Devin Desktop) | ~/.config/devin/mcp_config.json + stdio bridge |
| Claude Code | MCP remote exec + jobs on your fleet |
| Codex / ChatGPT | Same MCP endpoint or REST /api/v1/exec |
| Gemini / Grok | gemini mcp add / grok mcp add --transport http … |
How Zed handles MCP servers
- Zed uses the MCP tools and prompts features, and reacts to
notifications/tools/list_changed. Other MCP features are not used. - Servers go under
context_serversinsettings.json. For a remote server without anAuthorizationheader, Zed starts the standard OAuth flow on its own. - A green dot next to the server in the Agent Panel settings means it is running.
- From v0.224.0, approval is set with
agent.tool_permissions.default(confirm,allowordeny). Single MCP tools can be targeted withmcp:<server>:<tool>keys, for examplemcp:aicommander:remote_exec. - Custom agent profiles decide which tools a thread can use.
- Servers are also passed to external agents connected over ACP. Terminal threads run an agent's own CLI and use that CLI's MCP config instead.
Checked against Zed docs: Model Context Protocol in November 2026. Client behaviour changes between releases; the official docs win if they disagree.
A workflow that suits Zed
Zed's agent profiles let you build a narrow "remote ops" mode that can only touch your machines through AI Commander.
- Create a custom profile that enables the
aicommandertools and turns off the built-in terminal tool, so remote work can't end up running on your laptop by mistake. - Set
mcp:aicommander:remote_execandmcp:aicommander:remote_job_starttoconfirmand the read-only tools such asmcp:aicommander:list_machinestoallow. - In that profile, ask: "Check why nginx on the edge box returns 502s." The agent calls
remote_execfor logs and config, and each command waits for your confirmation.
Limits and gotchas in Zed
- Zed's docs note that how reliably tools get called depends on the model. If the agent ignores AI Commander, mention the server by name or switch models.
- Claude Code or Codex started in a terminal thread won't see servers from Zed's
context_servers. Add AI Commander to that CLI's own config. - Tool permission keys need Zed v0.224.0 or newer. Earlier versions used the
agent.always_allow_tool_actionsboolean instead.
FAQ
context_servers in Zed’s settings.json (Settings → AI → MCP Servers, or zed: open settings file): either a local server with command npx / args ["-y", "@aicommander/mcp"] and AICOMMANDER_SERVER=https://aicommander.dev, or a remote server with url https://aicommander.dev/mcp. The remote URL may ask for a one-time browser sign-in (OAuth); the npx bridge never does: it connects anonymously unless AICOMMANDER_TOKEN is set in its env. On the remote URL, use ?anonymous=1 for a code-only first hour. See /howto/ for the full matrix.mcp:aicommander:remote_exec and mcp:aicommander:remote_job_start to confirm, or set agent.tool_permissions.default to confirm (v0.224.0+).Next steps
Install the agent on a machine you control, connect Zed via context_servers, then ask for a real task on a remote box.