Keep GitHub Copilot.
Give it your machines.
AI Commander is not an alternative to GitHub Copilot, Claude Code, Codex, ChatGPT, Cursor, Grok, Gemini, Windsurf, Continue, Cline, Roo Code, Zed, OpenCode, Qwen Code, goose, Amp, or Kilo 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 · goose · Amp · Kilo Code · GitHub Copilot · Continue · Windsurf · Claude Code · Grok · Codex · ChatGPT · Amazon Q · Gemini.
Connect GitHub Copilot (MCP HTTP)
GitHub Copilot (Copilot CLI MCP docs) supports remote MCP over HTTP (Streamable HTTP) in Copilot CLI, VS Code Copilot Chat / agent mode, and the Copilot cloud agent. Point it at AI Commander’s hosted MCP endpoint:
Or persist under mcpServers in ~/.copilot/mcp-config.json
(same shape works for many Copilot agent MCP configs):
Optional Bearer header for account aliases:
Full client matrix (GitHub Copilot, Kilo Code, Amp, Claude Code, Codex, ChatGPT, Cursor, Grok, Gemini, OpenCode, Qwen Code, goose, Continue, Windsurf, Cline, Roo Code, Zed, REST): Connect your AI client. Copilot MCP docs: Add MCP servers (CLI) · Configure MCP (cloud agent).
First connect may open a one-time browser sign-in (OAuth).
For a code-only first hour without that flow, append ?anonymous=1 to the MCP URL.
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 Copilot 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 |
|---|---|
| GitHub Copilot | copilot mcp add --transport http aicommander https://aicommander.dev/mcp |
| Kilo Code | mcp.aicommander.url → https://aicommander.dev/mcp |
| Amp | amp mcp add aicommander https://aicommander.dev/mcp |
| goose | goose session --with-streamable-http-extension "https://aicommander.dev/mcp" |
| Qwen Code | qwen mcp add --transport http aicommander https://aicommander.dev/mcp |
| OpenCode | opencode mcp add aicommander --url https://aicommander.dev/mcp |
| Claude Code | claude mcp add --transport http … → remote exec + jobs |
| Codex / ChatGPT | Same MCP endpoint or REST /api/v1/exec |
| Cursor | npx @aicommander/mcp stdio bridge → remote exec + jobs |
| Grok / Gemini | grok mcp add --transport http … / gemini mcp add |
| Cline / Roo Code / Zed | MCP settings / context_servers + bridge or remote URL |
| Continue / Windsurf | stdio bridge in Continue / Windsurf MCP config |
How GitHub Copilot handles MCP servers
- Copilot CLI comes with the GitHub MCP server already set up. You add others with
copilot mcp add --transport http NAME URL, or from inside a session with/mcp add, where Ctrl+S saves the form. - The CLI stores servers in
~/.copilot/mcp-config.json. SetCOPILOT_HOMEto keep that directory somewhere else. - VS Code: run MCP: Add Server, or open Chat → Customizations → MCP Servers. A server saved to the "Copilot Global" scope goes into the same
mcp-config.json, so the CLI and the editor share it. - The older
.vscode/mcp.jsonuses aserverskey. Portable files (mcp-config.json, a workspace-root.mcp.json) usemcpServers. A block copied between the two needs that key renamed. - Portable configs don't support VS Code input variables or env files. A server that prompts for inputs is not passed on to Copilot's agent harness sessions.
- When the CLI asks to run a tool, you can allow it once or for the rest of the session.
--allow-all(alias--yolo) skips every prompt.
Checked against GitHub docs: Copilot CLI and VS Code docs: MCP servers in November 2026. Client behaviour changes between releases; the official docs win if they disagree.
A workflow that suits GitHub Copilot
- In a Copilot CLI session, ask it to start the training or migration run with
remote_job_startand note the job id. - Turn on experimental features with
/experimental on, then schedule a check-in such as/every 30m check AI Commander job <id> with remote_job_status and summarise the last log lines. Scheduled prompts only fire while that interactive session is running. - When the job ends, ask Copilot to pull the result file with
remote_pull(file courier, Pro) and open a pull request with the numbers.
Limits and gotchas in GitHub Copilot
- Approving
remote_exec"for the rest of the session" covers every machine and command in that session. Prefer approving once per command on production boxes. - Only use
--yolowhere an unexpected shell command on your machines would be acceptable. - If VS Code's agent never calls the tools, open the chat's tool picker and check that the
aicommandertools are enabled.
FAQ
~/.copilot/mcp-config.json. A server that only lives in .vscode/mcp.json is an editor-only setting.Next steps
Install the agent on a machine you control, connect GitHub Copilot via MCP HTTP to https://aicommander.dev/mcp, then ask for a real task on a remote box.