Windows Computer Use MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Windows Computer Use MCP Servertake a screenshot of my desktop"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
🖥️ Windows Computer Use MCP Server
A blazing-fast, lightweight, native Computer Use MCP server and plugin for Windows. Allows any AI model (Claude, GPT-4o, Gemini, DeepSeek, local LLMs) to interact with your local Windows desktop via mouse, keyboard, and screenshots.
Unlike Docker/Linux-only solutions, this runs directly on Windows with zero heavy dependencies (no Python, no PyAutoGUI, no Docker).
⚡ Sub-millisecond latency for mouse and keyboard actions, ~50 ms for screen capture.
✨ Features
🖱️ Mouse Control: Instant clicks (left, right, middle, double-click), cursor movements, and vertical scrolling.
⌨️ Native Unicode Typing: Fast text typing via Win32
SendInput(supports English, Russian, emojis, symbols regardless of keyboard layout).⌨️ Key Chords: Send hotkeys and special keys (
ENTER,TAB,ESC,CTRL+S,ALT+F4, etc.).📸 Fast Screenshot: Captures desktop to PNG in ~50ms.
⚡ Zero Heavy Runtime: Powered by Node.js and a pre-compiled native .NET C# daemon worker.
🔌 Universal Compatibility: Works with Claude Desktop, Antigravity, Cursor, Cline, Windsurf, or any MCP client.
Related MCP server: servo-mcp
⚡ Performance Benchmarks
Measured on a standard Windows machine:
Action | Latency |
| 0.8 ms |
| 0.5 ms |
| 1.6 ms |
| 26.4 ms (includes down/up physical click delay) |
| 31.5 ms |
| 41.1 ms |
| 62.8 ms (full capture + PNG encode + Base64 transmission) |
🛠️ Available Tools
Tool | Parameters | Description |
| (none) | Captures the current Windows desktop as a PNG image |
| (none) | Returns primary monitor resolution ( |
|
| Clicks at desktop coordinates |
|
| Moves pointer to |
|
| Types Unicode text into the focused control |
|
| Sends key or key chord (e.g. |
|
| Scrolls vertically (positive = up, negative = down) |
🚀 Quick Setup
Prerequisites
Windows 10 or 11
Node.js installed and in
PATH
Clone this repository:
git clone https://github.com/alikhanovdev-commits/windows-computer-use-mcp.git🔌 Connection Guides
1. Claude Desktop
Add to %APPDATA%\Claude\claude_desktop_config.json:
{
"mcpServers": {
"computer-use": {
"command": "node",
"args": ["C:\\path\\to\\windows-computer-use-mcp\\server\\server.mjs"]
}
}
}2. Google Antigravity & agy CLI
Merge into %USERPROFILE%\.gemini\config\mcp_config.json:
{
"mcpServers": {
"computer-use": {
"command": "node",
"args": ["C:\\path\\to\\windows-computer-use-mcp\\server\\server.mjs"]
}
}
}Or install as a plugin: copy this folder to %USERPROFILE%\.gemini\config\plugins\computer-use-windows.
3. Cursor / Windsurf / Cline (VS Code)
In your MCP settings, add:
Type:
stdioCommand:
nodeArgs:
C:\path\to\windows-computer-use-mcp\server\server.mjs
🛠️ Building From Source (Optional)
The pre-compiled worker.exe is already included. If you want to rebuild it:
build.bator via npm:
npm run build🔒 Security Notes
The server interacts with the active interactive desktop session running under your user account.
It will not bypass Windows UAC (Admin) elevation boundaries unless the host process is also run with elevated privileges.
Do not expose the server to untrusted remote networks without authentication.
📄 License
MIT License. Free for personal and commercial use.
This server cannot be deployed
Maintenance
Related MCP Connectors
Eyes and hands on real Windows PCs — observe, click, type via Glasswarp API.
Turns a phone into a camera+Bluetooth remote so AI assistants can see and control any PC.
Use your Mac, Windows or Linux computer from ChatGPT, Claude or Codex: files, commands, documents.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceA lightweight server that enables AI agents to interact natively with the Windows operating system for tasks like UI automation and application control. It allows LLMs to perform file navigation, simulate user input, and manage windows without requiring specialized computer vision models.MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to see and control your desktop with tools for screenshots, clicks, typing, and more, all locally on macOS and Windows.180 npmMIT
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to control Windows GUI applications like a human using screen capture, OCR, mouse and keyboard input, and window management, with safety levels and memory.-
- AlicenseNot gradedqualityAmaintenanceEnables AI assistants to autonomously control Windows 11 and 10 desktops via sub-10ms screen capture, native UI Automation element inspection, and zero-lag keyboard and mouse input. Combines a visual plane with a semantic plane and stall detection so agents can operate real applications reliably without vision-only guessing.13 npm34 PyPI4MIT