windows-desktop-use-mcp

An MCP server for controlling and perceiving Windows 11 from AI assistants. It provides AI with "eyes" (vision), "ears" (hearing), and "limbs" (input control), making the desktop environment accessible from MCP clients like Claude.

English | 日本語

Main Features

• Vision (Screen Capture): Capture monitors, specific windows, or arbitrary regions. GPU acceleration supported (enables capture of YouTube/Netflix without black screens).
• Hearing (Audio & Transcription): Record system audio or microphone, with high-quality local transcription using Whisper AI.
• Limbs (Desktop Input): Mouse movement, clicking, dragging, and safe navigation key operations (security restricted).
• Live Monitoring (Streaming): Monitor screen changes in real-time, viewable via HTTP streaming in a browser.
• Analysis: Structural text extraction from windows using UI Automation (Markdown) and AI-optimized visual features.

For Non-Developers (Pre-built .exe)

If you don't have a development environment, you can use the pre-built executable from Releases.

1. Download

1. Go to the Releases page
2. Download the latest WindowsDesktopUse.zip
3. Extract to your preferred location (e.g., C:\Tools\WindowsDesktopUse)

2. Configure Claude Desktop

Option A: Automatic setup

cd C:\Tools\WindowsDesktopUse
WindowsDesktopUse.exe setup

Option B: Manual setup Add this to your %AppData%\Roaming\Claude\claude_desktop_config.json:

{
  "mcpServers": {
    "windows-desktop-use": {
      "command": "C:\\Tools\\WindowsDesktopUse\\WindowsDesktopUse.exe",
      "args": ["--httpPort", "5000"]
    }
  }
}

3. Restart Claude Desktop

Close and reopen Claude Desktop to load the new MCP server.

For Developers (Build from Source)

1. Build

dotnet build src/WindowsDesktopUse.App/WindowsDesktopUse.App.csproj -c Release

2. Configure Claude Desktop

Option A: Automatic setup

WindowsDesktopUse.exe setup

Option B: Manual setup Add this to your %AppData%\Roaming\Claude\claude_desktop_config.json:

{
  "mcpServers": {
    "windows-desktop-use": {
      "command": "C:\\path\\to\\WindowsDesktopUse.exe",
      "args": ["--httpPort", "5000"]
    }
  }
}

3. Verify Installation

WindowsDesktopUse.exe doctor

CLI Commands

doctor - System Diagnostics

Check system compatibility and configuration.

WindowsDesktopUse.exe doctor
WindowsDesktopUse.exe doctor --verbose    # Show detailed information
WindowsDesktopUse.exe doctor --json       # Output in JSON format

setup - Claude Desktop Configuration

Automatically configure Claude Desktop integration.

WindowsDesktopUse.exe setup                              # Use default config path
WindowsDesktopUse.exe setup --config-path "C:\custom\path.json"  # Custom config path
WindowsDesktopUse.exe setup --no-merge                    # Overwrite existing config
WindowsDesktopUse.exe setup --dry-run                    # Show config without writing

whisper - Whisper AI Models

Manage Whisper AI models for audio transcription.

WindowsDesktopUse.exe whisper          # List available models and check installation
WindowsDesktopUse.exe whisper --list   # Show model list only

Available MCP Tools

Vision

• visual_list: List monitors, windows, or all. Switch with type parameter.
• visual_capture: Capture monitor, window, or region. Dynamic quality control (Normal=30/Detailed=70).
• visual_watch: Continuous monitoring/streaming. Switch video/monitor/unified with mode parameter.
• visual_stop: Stop all sessions (watch, capture, audio) with unified command.

Hearing

• listen: Record system audio or microphone and transcribe to text using Whisper AI.

Input Control

• input_mouse: Unified mouse operations (move, click, drag) with action parameter.
• input_window: Window operations (close, minimize, maximize, restore) with action parameter.
• keyboard_key: Press safe navigation keys (Enter, Tab, arrow keys, etc.). Text typing and modifier keys (Ctrl, Alt, Win) are blocked for security.

Utility

• read_window_text: Extract window text as Markdown using UI Automation.

Documentation Index

• Tools Reference - Detailed command list and usage examples (Japanese).
• Development Guide - Details on build, test, and architecture (Japanese).
• Whisper AI - Information about speech recognition features and models.
• Quality Test Report - Analysis of information quality differences by quality settings.

Requirements

• Windows 11 (or Windows 10 1809+)
• .NET 8.0 Runtime/SDK
• High DPI aware (Uses physical pixel coordinates)

License

MIT License. See LICENSE file.