Windows MCP Server

Windows system control and UI automation for LLMs across versions 7 through 11.

Data last scanned today · Reviewed today

50 4 dangerous tools20 tools★ 6,513PythonMIT

Overview

Windows MCP Server provides a bridge between AI agents and the Windows desktop environment. By acting as an interface for the operating system, it enables agents to perform tasks including file navigation, application control, and UI interaction. The server operates without reliance on traditional computer vision, allowing it to function with any LLM. It includes built-in tools for mouse and keyboard simulation, window state capturing, and general UI manipulation on Windows 7, 8, 10, and 11.

Our verdict

With a security score of 50/100, this project is classified as high-risk. This rating is driven by the fact that it is a community-maintained tool that has not yet been editorially reviewed or had its authentication and read-only modes formally assessed. While it benefits from having zero direct dependencies, users should exercise caution as it grants AI agents broad control over their desktop environment.

Supports all Windows versions from 7 to 11
Requires zero direct dependencies
Compatible with any LLM without dedicated vision requirements
Licensed under the permissive MIT license

Community-maintained rather than official vendor software
Authentication methods remain unreviewed
Read-only mode implementation is not yet verified

Tools

Tool	Description	Risk
Click	Click on the screen at the given coordinates.	medium
Type	Type text on an element (optionally clears existing text).	medium
Scroll	Scroll vertically or horizontally on the window or specific regions.	medium
Move	Move mouse pointer or drag (set drag=True) to coordinates.	medium
Shortcut	Press keyboard shortcuts (Ctrl+c, Alt+Tab, etc).	medium
Wait	Pause for a defined duration.	low
WaitFor	Wait until text, an active window, an element, or a focused element appears by polling UI state inside one tool call.	low
DisplayInventory	Read display layout, work areas, effective DPI, and scale metadata.	low
Screenshot	Fast screenshot-first desktop capture with cursor position, active/open windows, and an image.	low
Snapshot	Full desktop state capture for workflows that need interactive element ids, scrollable regions, or use_dom=True browser extraction.	low

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What are the system requirements for Windows-MCP?

You need a machine running Windows 7 through 11 with Python 3.13+ and the UV package manager installed.

›Are there any specific language settings I should be aware of?

Yes, English is the preferred default language for Windows. If you use a different system language, you should disable the App-Tool to avoid issues.

›Can this be configured to run automatically?

Yes, you can run the 'windows-mcp install' command to create a background scheduled task that starts the server at every login.

›How can I troubleshoot the server or check for errors?

Logs are stored in the '~/.windows-mcp/' directory, specifically within 'server.log' and 'server.error.log'.

›What is the expected behavior during the first installation?

The initial run may take a minute or two to install dependencies defined in pyproject.toml, and the server might time out on its first start; simply ignore this and restart it.

›Does the browser automation mode support all browsers?

The DOM mode supports Chrome, Edge, and Firefox, though Firefox uses an IAccessible2 fallback because it does not expose RootWebArea via UIA.