Summary

This project provides a Model Context Protocol (MCP) server enabling AI models, particularly Claude, to control a user's computer. It offers an easy-to-set-up solution for developers and power users experimenting with AI-driven automation, granting AI models direct operational capabilities on a local machine, albeit with significant security caveats.

How It Works

The computer-use-mcp server implements a computer interaction tool using nut.js for system control, inspired by Anthropic's official guide. It acts as an MCP server, allowing AI models to send commands and receive feedback. A key design choice nudges models to prioritize keyboard shortcuts over coordinate-based actions for increased UI automation reliability.

Quick Start & Requirements

Installation is streamlined across Claude, Cursor, and Cline:

• Claude: Use claude mcp add --scope user --transport stdio computer-use -- npx -y computer-use-mcp for command-line installation, or download .mcpb artifacts from GitHub Actions for manual installation via Claude Desktop. Advanced users can configure claude_desktop_config.json.
• Cursor: A one-click install option is available via a deep link, or via JSON configuration in ~/.cursor/mcp.json or .cursor/mcp.json.
• Cline: Install directly from the marketplace or configure via JSON within the Cline extension. Node.js is a prerequisite for advanced JSON configuration methods. For optimal performance, using models adept at computer interaction (like latest Claude models), a 720p resolution, and the Rango browser extension are recommended.
• GitHub Actions artifacts: https://github.com/domdomegg/computer-use-mcp/actions/workflows/ci.yaml?query=branch%3Amaster
• Cursor one-click install: https://cursor.com/install-mcp?name=computer-use&config=JTdCJTIyY29tbWFuZCUyMiUzQSUyMm5weCUyMC15JTIwY29tcHV0ZXItdXNlLW1jcCUyMiU3RA%3D%3D

Highlighted Details

• Grants AI models "complete control" over the host computer.
• Designed for straightforward local setup and integration.
• Emphasizes keyboard shortcut usage for AI interactions.
• Demonstrated capability includes AI-driven desktop background changes.

Maintenance & Community

Contributions are welcomed via GitHub pull requests. Development uses standard Node.js tooling (npm install, npm run test, npm run build). Releases follow semantic versioning and are published via GitHub Actions to the NPM registry. Specific community channels like Discord or Slack are not detailed in the README.

Licensing & Compatibility

The license type is not explicitly stated in the provided README. This omission requires clarification for assessing commercial use or closed-source linking compatibility.

Limitations & Caveats

This project carries significant risks due to the AI's complete system control. Models are prone to errors and prompt injection vulnerabilities, potentially causing substantial damage. Users must supervise AI operations closely, consider sandboxed environments, and exercise extreme caution. Reliability is enhanced by tools like the Rango browser extension.