SkillFoxx

Type a name, a task or a service

Desktop Commander MCP

An MCP server that lets an agent work on your computer: terminal and long-running processes, file search and edits, Excel, PDF and DOCX

MCP serverEditors’ pickHigh risk

wonderwhy-er/desktopcommandermcpAuthor: Eduard Ruzga

Install

npx @wonderwhy-er/desktop-commander@latest setup

Configures Claude Desktop; restart the app afterwards.

This is third-party code. Review the repository files before installing.

What it does

Desktop Commander gives the agent a terminal with streaming output, background runs and process control, including interactive ones like SSH and database shells. File tools read and write files, search content with ripgrep and make targeted code replacements. It also handles Excel, PDF and DOCX, and can run Python, Node.js and R code in memory without saving files. Tool calls are logged locally, and settings let you define allowed directories and blocked commands.

Who it is for. For Claude Desktop and other chat client users who want the model to work with files and the terminal on their machine.

Good fit when

  • You want Claude Desktop to access the terminal and files
  • You need to analyze CSV or Excel without extra tools
  • You need to run long commands and watch their output

Not a fit when

  • An agent like Claude Code or Codex already has its own terminal and file tools
  • The machine holds data the model must not access

Example request

Open sales.xlsx in Documents, sum revenue by month and save the result to a new sheet

Limitations

Safeguards are not a sandbox: directory limits do not stop terminal commands from reaching other files, and the authors suggest Docker for full isolation. Optional telemetry exists and is turned off with the telemetryEnabled setting. Running via npx with @latest auto-updates the server, which is convenient but changes code without your involvement.

How to disable. Run npx @wonderwhy-er/desktop-commander@latest remove, or claude mcp remove desktop-commander, or delete the entry from the client config.

MCP

Transport
stdio, http
Authentication
not required

Security check

  • Runs arbitrary shell commands on your machine
  • Reads, edits and deletes files outside the project

README in short

The README lists capabilities: terminal and processes, file operations, Excel, PDF and DOCX, in-memory code execution, a local call log and safety measures. It documents six install options for Claude Desktop, including npx, a script, Smithery, manual config, a clone and Docker, with auto-update notes. It gives commands for Claude Code, Codex and other clients, a remote mode for ChatGPT and Claude on the web, and a separate desktop app. MIT licensed.

FAQ

Can I restrict folder access?

Yes, via allowedDirectories, but the README warns that terminal commands bypass this limit.

Are my commands sent to a server?

Call history is stored locally. Remote mode for web clients goes through a cloud service that, per the authors, does not retain arguments or results.

Editors’ pick

A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more

CLIHigh risk31K
Editors’ pick

Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server

CLIMedium risk29.7K
Editors’ pick

A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI

MCP serverMedium risk28.1K
Editors’ pick

Anthropic's official plugins for knowledge work roles: sales, support, product, marketing, legal, finance, data and enterprise search

PluginHigh risk24K