Windows MCP

by CursorTouch

Windows MCP is an open-source MCP server that bridges AI agents with the Windows operating system. Built by CursorTouch, it reached 2M+ users through the Claude Desktop Extensions directory and has 5,100 GitHub stars. The key differentiator: it doesn't use computer vision or fine-tuned models. Any LLM works. Connect Claude, GPT, Gemini, or any MCP-compatible client, and the server gives your AI direct access to Windows UI elements — clicking buttons, typing text, scrolling, dragging, taking screenshots, and controlling applications. Available tools include Click, Type, Scroll, Move (with drag), Keyboard Shortcuts, Wait, and Screenshot (with fast capture and multi-display support). A DOM mode (use_dom=True) enables streamlined browser automation without relying on screen coordinates. Response latency runs 0.2-0.9 seconds between actions. Runs on Windows 7 through 11. Requires Python 3.13+ and the UV package manager. Install with uvx windows-mcp or clone from source. Supports both local mode (direct Windows automation) and remote mode (proxy connection to windowsmcp.io for cloud-hosted VM automation). Compatible clients: Claude Desktop, Perplexity Desktop, Gemini CLI, Qwen Code, Claude Code, and other MCP-compatible applications. MIT licensed. For related AI automation tools, browse our AI tools directory or check out other MCP servers and repositories for extending your AI agent capabilities.

FAQ