MCPComputerUse

MCPComputerUse - C# Computer Use MCP Server

A native C# implementation of a Computer Use MCP Server for Windows, providing superior automation capabilities through direct Windows API integration.

- Screenshot Capture: Multi-monitor support with flexible targeting
- Window Management: Complete window enumeration and control
- Mouse Automation: Precise clicking, moving, and scrolling
- Keyboard Automation: Text typing and key combinations
- Macro System: Complex automation sequences
- MCP Protocol: Full compliance with MCP specification
- High Performance: Native Windows API integration
- Zero Dependencies: Self-contained executable

- computer-use:take_screenshot- Take screenshots with flexible targeting
- Support for multi-monitor setups
- Window-specific captures
- Configurable output formats

- computer-use:list_windows- List all visible windows
- computer-use:get_active_window- Get currently active window info
- computer-use:focus_window- Focus a window by ID or name
- computer-use:manage_window- Minimize, maximize, restore windows

- computer-use:mouse_click- Click at coordinates with button options
- computer-use:mouse_move- Move mouse to coordinates
- computer-use:type_text- Type text with Unicode support
- computer-use:press_key- Press keys with modifier support
- computer-use:scroll- Scroll at specific coordinates

- computer-use:run_macro- Execute complex automation sequences
- Support for all automation actions
- Error handling and reporting
- Configurable timing

- computer-use:get_server_capabilities- Get server capabilities and status

- MCPComputerUse.exe- Main console application with MCP protocol
- WindowsAutomation.dll- Native Windows API automation library
- MCPProtocol.dll- MCP specification implementation

- .NET 8.0- Latest LTS framework
- Windows Forms- For screen capture support
- System.Drawing- Image manipulation
- Newtonsoft.Json- JSON serialization
- Native Windows APIs- Direct P/Invoke integration

- Windows 10 or Windows 11
- No additional runtime required (self-contained)
- Download the latest release from GitHub
- ExtractMCPComputerUse.exeto your desired location
- No installation required - single executable file

Add to your Claude Desktop configuration:

{ "mcpServers": { "computer-use": { "command": "c:/path/to/MCPComputerUse.exe", "args": [] } } }

- Visual Studio 2022 or VS Code with C# extension
- .NET 8.0 SDK
- Windows 10/11 development machine

# Clone the repository git clone <repository-url> cd MCPComputerUse # Restore dependencies dotnet restore # Build solution dotnet build --configuration Release # Publish self-contained executable dotnet publish src/MCPComputerUse -c Release -r win-x64 --self-contained -o ./publish

The resulting executable will be in the./publishfolder.

{ "method": "tools/call", "params": { "name": "computer-use:take_screenshot", "arguments": { "target": "screen", "filename": "desktop_capture" } } }

{ "method": "tools/call", "params": { "name": "computer-use:mouse_click", "arguments": { "x": 500, "y": 300, "button": "left", "clicks": 2 } } }

{ "method": "tools/call", "params": { "name": "computer-use:run_macro", "arguments": { "commands": [ { "action": "click", "x": 100, "y": 100 }, { "action": "type", "text": "Hello World" }, { "action": "key", "key": "enter" } ] } } }

- Ensure the executable has permission to capture screens
- Some applications may require running as administrator

- UAC-protected applications may not respond to focus commands
- Try using window IDs instead of window titles for better reliability

- Check display index for multi-monitor setups
- Verify target window exists and is visible

The application logs to console output. Use the following log levels:

- Information: General operation status
- Debug: Detailed execution information
- Error: Error conditions and failures

- Screenshot capture: <100ms for 1920x1080
- Window enumeration: <50ms for 100 windows
- Mouse/keyboard actions: <10ms per action
- Macro execution: 100ms between commands (configurable)

- Base memory: <50MB
- Per screenshot: <10MB temporary
- Window caching: <5MB

- No elevation required for basic operations
- Screen capture permissions may be requested by Windows
- Input simulation works with most applications

- Coordinate validation prevents out-of-bounds operations
- Rate limiting prevents input flooding
- Emergency stop sequences available

Contributions are welcome! Please follow these guidelines:
- Fork the repository
- Create a feature branch
- Write tests for new functionality
- Ensure all tests pass
- Submit a pull request

# Clone your fork git clone <your-fork-url> cd MCPComputerUse # Install dependencies dotnet restore # Run tests dotnet test # Start development code .

- OCR Integration- Text recognition from screenshots
- Image Matching- Find UI elements by template
- Macro Recording- Record user actions automatically
- Plugin System- Extensible architecture
- UI Automation- Windows accessibility API integration

- Async Operations- Better responsiveness
- Caching- Reduced API calls
- Optimization- Faster screenshot capture

This project is licensed under the MIT License - see theLICENSEfile for details.

MCP allowing your Agent to request human validation on webapps and apps.