Desktop Commander is a local MCP server that lets AI assistants run terminal commands, manage processes, edit files, and execute code in memory. It runs on a user's local machine using existing host client subscriptions and supports optional cloud-based remote MCP connections.
Project overview
The project enables AI assistants to execute long-running terminal commands, manage interactive processes, and perform in-memory data analysis without saving files, operating through the user's existing host client subscription rather than API token costs.
Project type
MCP
Deployment
Refer to project documentation
License
MIT
Best for
Developers using MCP-compatible AI clients who need to execute terminal commands, manage local processes, and perform in-memory code analysis on their local machine.
Users who prefer to operate within their existing host client subscriptions rather than relying on separate API token costs for AI-driven local tasks.
Key capabilities
Execute long-running terminal commands with output streaming, manage interactive processes such as SSH or dev servers, and handle session management.
Execute code in Python, Node.js, and R directly in memory without saving files for instant data analysis.
Use Desktop Commander from ChatGPT, Claude web, and other AI services via a cloud Remote MCP service connected to a local device.
Limitations and risks
The allowedDirectories setting currently only restricts filesystem operations and does not restrict terminal commands.
Directory restrictions and command blocking can be bypassed through symlinks, command substitution, and absolute paths.
The AI may attempt to modify configuration settings such as allowedDirectories if it encounters filesystem access restrictions.
Getting started
Setup is rated as easy due to automated one-line installation scripts and multiple installation methods that do not require manual coding.
Run the automated setup command and then restart the MCP client to begin using the server.
Alternatives and comparisons
A CLI tool that brings Gemini into the terminal for code understanding, generation, and workflow automation with MCP support, providing an alternative terminal-based AI approach.
A provider-agnostic, local-first AI agent CLI that executes shell commands, writes code, and browses the web directly from the terminal.