Mac Use Mcp

Gives AI agents full desktop control on macOS through 18 tools: screenshot capture, mouse clicks and drags, keyboard input with Unicode support, window management, and accessibility tree inspection. Zero native dependencies, runs via npx with no build step. Requires macOS Accessibility and Screen Recording permissions. Useful for automating UI tests, scraping visual data from desktop apps, or building agents that need to interact with any macOS application through vision and input rather than APIs. The screenshot→action loop can rack up tokens quickly, so pick a fast model with solid vision capabilities. Works on macOS 13+ on both Intel and Apple Silicon.