Desktop Commander is an MCP server that hands your AI assistant the keys to your computer: terminal control, process management, deep file search and precise diff-based editing. Built for Claude and other MCP clients, it lets an agent actually do work across your projects instead of only suggesting code. It has about 10,100 stars on GitHub and is actively developed in 2026.
Installed in seconds via npx, Desktop Commander exposes a rich toolset over the Model Context Protocol: run and interact with terminal commands and long-lived processes, search file contents across whole directory trees, read and write files with surgical diff edits that change only the intended lines, and analyze code structure across multiple projects at once. Because it rides on your existing Claude subscription through Claude Desktop or other MCP hosts rather than raw API calls, heavy agent sessions do not rack up per-token API bills. Security controls include command allowlists and directory restrictions so you decide exactly where the agent may roam, making it a pragmatic bridge between chat assistants and real desktop automation. The server is cross-platform across Windows, macOS and Linux, and updates itself when installed through npx, so new capabilities arrive without manual maintenance. Its toolset grew out of real agentic coding sessions: reading specific line ranges, moving files and managing long-running processes reflect what agents actually need on a desktop. The same team is also developing a companion desktop application for users who prefer a graphical setup over editing MCP configuration files.
Desktop Commander MCP is free and open source under the MIT license. You only need an account with the AI client you connect it to.
Desktop Commander is for developers and power users who want their AI assistant to execute real work โ editing code, running builds, organizing files โ across the whole machine, not just one editor.