Your connection, step by step.
RemoteCommand uses the same Node.js agent on Windows, macOS and Linux. Node.js LTS is recommended; npx runs the npm package without a separate manual install.
Sign in to RemoteCommand
Once sign-in is enabled, continue with Google on the authentication page. Existing Google-owned email accounts can reconnect to their existing account after verification.
Open sign-in →Register your agent
Install Node.js LTS, then run this one command in Windows PowerShell or macOS/Linux Terminal. There is no separate package installation or enrollment token to paste.
npx -y @lhtsoft/remotecommand-agent@latest
Is npx missing or not recognized?
Install the current Node.js LTS from the official site. Node.js normally includes npm and npx; you do not need to install npx separately. Get Node.js LTS
Close and reopen your terminal after installing Node.js. Then check:
node -v npm -v npx -v
If npm is available but npx is not, check your Node.js/npm installation and PATH. The -y option skips npm's install confirmation; @latest selects the newest published stable version. Neither updates an already running agent.
Your browser opens for sign-in and device approval. Select the folders and permission profile there. Keep the terminal running while you want this device connected.
To update later, finish current work, stop the agent with Ctrl+C and run the latest version with your saved registration:
npx -y @lhtsoft/remotecommand-agent@latest run
Use the same OS account as your first registration. Reinstalling with npx does not remove your device ID or saved permissions. Do not run a second agent for the same device.
Connect your AI client
Install RemoteCommand from the official ChatGPT Plugins listing, then sign in and approve device access. Claude integration is coming soon; for other MCP clients, use the manual guide below.
Open RemoteCommand in ChatGPT ↗
Other AI apps · Manual MCP connection
For ChatGPT, use the official RemoteCommand plugin above. Claude integration is coming soon. Use this manual MCP setup for other compatible apps.
- In Cursor, open Settings → MCP and add a remote MCP server. You can also place the configuration below in ~/.cursor/mcp.json.
- Enable the server and complete the RemoteCommand OAuth sign-in and permission approval.
- Ask Cursor to run list_devices to confirm your computer is online.
{
"mcpServers": {
"remotecommand": {
"url": "https://api.remotecommand.app/mcp"
}
}
}Official Cursor documentation ↗Your AI app needs to support remote MCP and OAuth. Device access permissions still apply.
AI Efficient Mode
Send less noise to AI. Compress noise, not information.
Your dashboard shows measured output data reduction by device and period. Smaller tool responses can save AI input tokens; supported agents also estimate saved tokens from actual text, not billing data.
Reduces unnecessary logs and generated output sent to AI assistants while preserving access to requested source code, documents, and other original content. Large content is delivered in ranges instead of being silently discarded.
Open your device page to configure Smart/Tail/Full output, line and character limits, error extraction and local log storage. New agents default to ON; existing configurations default to OFF. Full output remains subject to transport limits.
search_files → read_file range → next range if hasMore → apply_patch → run_tests → git_diff
For Docker/log analysis, request output: {mode: "tail", maxLines: 100}. The dashboard measures output reduction in bytes and separately estimates saved input tokens for tokenizer-enabled agent runs; neither metric is a direct billing estimate.
Start with a read-only request
Try: “List my connected computers, then show the files in my allowed project folder.” Enable terminal access separately when needed. Review activity in the dashboard and on your agent.
Open device management →