Keep Cursor.
Give it your machines.
AI Commander is not an alternative to Cursor, Grok, Claude Code, Codex, or ChatGPT. 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.
Sibling harness pages: Claude Code · Cursor · Grok · Codex · ChatGPT.
Connect Cursor (stdio bridge)
Cursor talks to AI Commander through the stdio MCP bridge. Add this to
~/.cursor/mcp.json (or Settings → MCP → Add server):
{
{
"mcpServers": {
"aicommander": {
"command": "npx",
"args": ["-y", "@aicommander/mcp"],
"env": {
"AICOMMANDER_SERVER": "https://aicommander.dev"
}
}
}
}
HTTP clients can use https://aicommander.dev/mcp directly.
Full client matrix (Grok, Claude Code, Codex, ChatGPT, Cursor, REST):
Connect your AI client.
First connect may open a one-time browser sign-in (OAuth Dynamic Client Registration).
For a code-only first hour without that flow, append ?anonymous=1 to the MCP URL 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 Cursor 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 |
|---|---|
| Cursor | npx @aicommander/mcp stdio bridge → remote exec + jobs |
Grok / grok CLI | grok mcp add --transport http … → remote exec + jobs |
| Claude Code | MCP remote exec + jobs on your fleet |
| Codex | Same MCP endpoint; aliases after sign-in |
| ChatGPT | Custom MCP connector or plain REST /api/v1/exec |
FAQ
npx -y @aicommander/mcp and AICOMMANDER_SERVER=https://aicommander.dev. First connect may open a one-time browser sign-in (DCR). Use ?anonymous=1 for a code-only first hour. HTTP clients can also use https://aicommander.dev/mcp — see /howto/ for the full matrix.Next steps
Install the agent on a machine you control, connect Cursor via mcp.json, then ask for a real task on a remote box.