Harusame64/desktop-touch-mcp

What it does

Harusame64/desktop-touch-mcp is a Model Context Protocol (MCP) server that enables LLM agents to automate Windows desktop interactions. Unlike traditional coordinate-based clicking, it uses entity-based actions to identify and interact with UI elements, making interactions more reliable and context-aware. The tool integrates multiple automation methods including UI Automation (UIA), Chrome DevTools Protocol (CDP), screenshot analysis, keyboard/mouse/clipboard control, and terminal command execution.

How to set up

This is an MCP server designed for integration with compatible AI platforms. Installation typically involves: 1) Clone or download the repository from GitHub, 2) Configure the MCP server settings in your AI platform (Claude, ChatGPT with MCP support, or AI Studio), 3) Establish connection between the MCP server and your LLM client, 4) Grant necessary Windows permissions for desktop automation and terminal access. Detailed setup instructions are available in the project's README documentation.

Key features

• Entity-based UI interaction instead of coordinate-only clicking
• Multi-method automation: UIA, CDP, screenshots, keyboard, mouse, clipboard, and terminal control
• Entity leases for managing long-running operations
• Verified delivery confirmation for critical actions
• Causal context tracking for action sequences
• Windows desktop and application automation for LLM agents