How it works
Install the RemoteCommand agent
Install a supported Node.js LTS release (22, 24 or 26), run the agent locally, and register your computer. Choose allowed folders and an appropriate permission profile.
npx -y @lhtsoft/remotecommand-agent@latestOpen RemoteCommand in ChatGPT
Use the official ChatGPT Plugin Directory listing rather than manually creating a developer-mode MCP connector. Add RemoteCommand, select your RemoteCommand account, and review OAuth permissions.
Verify device access
Keep the agent running and connected. In a ChatGPT conversation with RemoteCommand enabled, request list_devices. Actual file, terminal, Git and Docker actions remain subject to the agent policy and approvals.
ChatGPT and RemoteCommand accounts
Your ChatGPT login can differ from your RemoteCommand account. Sign in with the RemoteCommand Google account that owns the registered computer. New RemoteCommand accounts may require an invitation while the service is in Closed Beta.
Permissions and safety
Access is not automatic after adding the plugin. You must authorize the connection, and the agent enforces your folder, command, deletion, Docker and approval policies. Grant only the access needed for your task.
Frequently asked questions
Can I install RemoteCommand from ChatGPT now?
Yes. The official RemoteCommand listing is available from the ChatGPT Plugin Directory. Access may depend on your ChatGPT plan and workspace, and RemoteCommand account eligibility is checked separately.
Do I need developer mode or a custom MCP URL?
No. Use the official plugin listing. The manual Streamable HTTP MCP setup remains available for other compatible clients.
Why are my devices not listed after adding the plugin?
Sign in to the correct RemoteCommand account, start your registered local agent, and verify its connection. Device access still follows your local policy.