@wonderwhy-er/desktop-commander

What it does

Desktop Commander exposes file operations, terminal commands, and code execution to Claude via MCP. It reads and writes files across your filesystem, executes terminal commands with output streaming, manages running processes, and works with structured formats (Excel, PDF, DOCX). Code runs in-memory (Python, Node.js, R) without disk writes. The server includes recursive directory search, audit logging, process output pagination, and symlink traversal protection.

Who it's for

Developers and DevOps engineers who want Claude to interact directly with their local system—searching codebases, editing files in bulk, running build commands, analyzing data files, and managing services from a single chat interface without juggling separate tools.

Common use cases

• Execute long-running terminal commands and manage processes directly from Claude
• Search codebases and perform bulk file edits with AI-driven regex replacements
• Analyze CSV, JSON, and Excel data files interactively in chat
• Automate local development tasks (builds, tests, deployments) triggered from conversation
• Read and modify Word documents, PDFs, and other structured files programmatically

Setup pitfalls

• Default install via npx requires Node.js; Docker option available for isolated, containerized environments
• High risk class: grants Claude filesystem read/write and arbitrary command execution—verify the command blocklist before connecting to untrusted AI instances or shared systems
• Long-running commands can exceed Claude's context window; use output pagination with offset/length to manage large outputs
• Verify path restrictions if multiple users or agents share the same system