Best for
- Developers and testers who want to delegate Windows desktop automation to AI in a controlled environment
- Researchers and AI experimenters needing programmatic desktop control
- Users building AI agents that require visual feedback
M
44FMRS · D
MCPControl is an MCP server for Windows OS automation, providing programmatic control over mouse, keyboard, window management, and screen capture. It bridges AI models with your desktop, enabling secure control of mouse movements, clicks, keyboard input, window operations, screenshots, and clipboard. It supports Windows only and recommends a virtual machine at 1280x720 resolution for optimal performance.
Tools
The tool list has not been reviewed yet.
Setup
- Install Visual Studio 2022 Build Tools (including VC++ workload).
- Install Python 3.12 (required for node-gyp).
- Install Node.js (latest LTS).
- Install MCPControl globally:
npm install -g mcp-control. - Start the server with SSE transport:
mcp-control --sse. - Configure your client with the SSE URL (e.g., http://192.168.1.100:3232/mcp).
Fit and risk
Not for
- Non-Windows systems (Windows only)
- Multi-monitor or non-standard resolutions requiring high click accuracy
- Production environments with high security requirements without proper safeguards
Required permissions
- Mouse control
- Keyboard control
- Window management
- Screen capture
- Clipboard access
Risks and side effects
- This software is experimental and may cause unintended consequences.
- Direct control of mouse and keyboard can lead to unintended actions or data loss.
- HTTPS is mandatory for remote deployments to avoid interception.
- Some operations may require elevated permissions.
Troubleshooting
- Ensure Visual Studio Build Tools and Python are installed for native modules.
- Check that the server is running with SSE mode and the network port is accessible.
- For remote access, use HTTPS with valid TLS certificates.
- Verify the URL in the client configuration is correct.
- Review known limitations, such as unsupported minimize/restore and get_screenshot not working in Cline.
Use cases
Delegate repetitive UI testing to AI
Automate workflows and form filling
AI gaming and visual reasoning experiments
Cross-application integration and demo creation
Supported clients
Supported clients have not been confirmed yet.