Remote Desktop Commander MCP
Hosted MCP server that links claude.ai, ChatGPT and Cursor to your computer's files and terminal over OAuth
- 222 ★ on the repository
- Last commit
Broad access
Why this level
- The agent gets full access to your machine's files and shell with your account's permissions, including deletion and running processes
- The connection goes through a third-party relay, so connected AI accounts are worth protecting with MFA and unused devices revoked
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/remote-desktop-commanderDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent gets full access to your machine's files and shell with your account's permissions, including deletion and running processes
- The connection goes through a third-party relay, so connected AI accounts are worth protecting with MFA and unused devices revoked
This is third-party code. Review the repository files before installing.
What it does
Remote Desktop Commander is a hosted MCP server that relays a secure connection to your real computer. An AI app gets the same file system and terminal access as the local Desktop Commander without you being at the keyboard. On the target computer you start the device agent with `npx @wonderwhy-er/desktop-commander@latest remote`, the browser opens a page with a confirmation code and after sign-in the terminal reports Device ready. The server is at `https://mcp.desktopcommander.app/mcp` (Streamable HTTP, OAuth 2.0 with PKCE). The tools read, search, write and edit files, start and stop processes, list paired devices and manage settings. Several computers can be paired to one account and addressed by name.
Who it is for. People who work in claude.ai, ChatGPT or Cursor and want to drive their machine's files and terminal from anywhere without sitting at it.
Good fit when
- From a phone you need to start a build on a home computer and read the process output
- The agent edits a file on a remote machine and reads a log while you are on the road
- Several paired computers are used in one chat and the agent addresses each by name
- Word, Excel and PDF documents are edited on disk next to the project
Not a fit when
- Work stays on this same computer: the local Desktop Commander is simpler, open source and needs no relay
- The machine cannot keep the device agent running: there is no access without it
- Data must not pass through a third-party relay: the connection goes through mcp.desktopcommander.app
Example request
On my home computer open the project folder, run the tests and show the last twenty lines of outputMCP
- Transport
- http
- Authentication
- OAuth
Limitations
The server is hosted, its code is closed and proprietary, and the repository holds only manifests, documentation and an issue tracker. The service is in beta. Tools run with your user account's permissions on the paired machine, so connected AI accounts are worth protecting with MFA. Access exists only while the device agent runs on `Node.js 18` or newer: stopping it with `Ctrl+C` ends access and devices are revoked in the dashboard. macOS, Windows and Linux are supported. The open-source local server lives in the DesktopCommanderMCP repository.
How to disable. Stop the device agent (`Ctrl+C`), revoke the device in the mcp.desktopcommander.app dashboard and remove the server from the client (`claude mcp remove remote-desktop-commander`).
FAQ
How do I connect Remote Desktop Commander to Claude Code?
First start the device agent on your computer, then run `claude mcp add --transport http remote-desktop-commander https://mcp.desktopcommander.app/mcp` and sign in over OAuth.
Does the terminal have to stay open?
Yes, the device agent runs in a terminal and the machine is reachable while it runs. Closing the terminal or `Ctrl+C` ends access.
How is it different from the local Desktop Commander?
The local server works over stdio on one machine and is open source. The remote one is a hosted relay that gives the same access from claude.ai, ChatGPT and mobile apps.
How does sign-in stay safe?
OAuth 2.0 with PKCE, and devices pair through the device authorization flow with a short-code check. Transport uses HTTPS and TLS and the server gets only data requested by tool calls.
Can I connect several computers?
Yes, several computers are paired to one account and addressed by name in one conversation.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server that gives Claude Desktop and other clients file tools, a shell and sub-agents in the style of Claude Code
A command-line debugger for MCP servers from Giant Swarm: a REPL for tools, OAuth 2.1 and an MCP server mode for Cursor
An MCP server that lets an agent read files, search code and run commands in a workspace
MCP server that turns a RAD Studio machine into a remote backend for agents: read, safe editing, MSBuild builds, git and deploy
A self-hosted MCP server that gives ChatGPT and other clients access to code on your machine: read, edit, search and run commands
An MCP bridge that lets ChatGPT operate your Mac: shell, files, terminal sessions, Codex history and background Chrome.
README in short
The text covers remote access to a computer from AI apps, a two-step quick start, setup for Claude, ChatGPT, Cursor and VS Code clients, a tool table and the relay design. Support and legal documents follow.