computer-use-mcp and mcp-vnc
These are **competitors** — both provide remote control interfaces for AI agents to operate a desktop, with the former using direct system APIs and the latter using VNC protocol, serving the same use case of giving language models computer automation capabilities.
About computer-use-mcp
domdomegg/computer-use-mcp
💻 Give AI models complete control of your computer (probably a bad idea)
Implements the Model Context Protocol to expose computer control via screenshot capture, mouse/keyboard input, and OCR—mirroring Anthropic's computer use API but optimized for local deployment. Built on nut.js for cross-platform input automation, with keyboard-first command preferences and optional Rango browser extension integration for reliable web navigation. Integrates directly with Claude Desktop, Claude Code, Cursor, and Cline via stdio transport configuration.
About mcp-vnc
hrrrsn/mcp-vnc
Remote desktop control for AI
Implements the Model Context Protocol (MCP) to expose VNC remote desktop control as tools for AI agents, supporting mouse clicks, keyboard input, text typing, and screenshot capture across Windows, Linux, and macOS. Integrates via stdio transport with Claude Desktop and VS Code's Copilot, connecting to standard VNC servers through configurable host/port/password credentials.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work