CursorTouch/Windows-MCP

What it solves

Windows-MCP provides a bridge between AI agents (LLMs) and the Windows operating system. It allows AI agents to interact with the OS natively, enabling them to perform tasks like navigating files, controlling applications, interacting with the UI, and conducting QA testing without requiring specialized computer vision models.

How it works

It operates as a Model Context Protocol (MCP) server that exposes a set of tools to a connected AI client. It interacts with Windows UI elements and simulates user input (keyboard and mouse) to control the system. It also includes a specialized "DOM Mode" for browser automation that filters out browser UI elements to focus exclusively on web page content for Chrome, Edge, and Firefox.

Who it’s for

Developers and QA engineers who want to integrate LLMs with Windows automation, as well as users of MCP-compatible clients (like Claude Desktop, Perplexity, and Gemini CLI) who want their AI agents to have direct control over their Windows environment.

Highlights

• Native Windows Integration: Controls windows, opens apps, and simulates user input across Windows 7 through 11.
• LMM Agnostic: Works with any LLM and does not rely on traditional computer vision or fine-tuned models.
• Browser Automation: Features a DOM mode for cleaner web automation by filtering browser UI.
• Flexible Transport: Supports stdio, SSE, and streamable-HTTP for both local and network-based access.
• Enterprise Security: Includes authentication, IP allowlists, CORS origin control, TLS/HTTPS, and OAuth 2.0 + PKCE support.
• Low Latency: Typical action latency ranges between 0.2 to 0.5 seconds.