MCP Remote macOS Use: AI Agent for Full macOS Control via Claude Desktop

Introduction

MCP Remote macOS Use is an innovative open-source MCP server designed to empower AI agents with full control over remote macOS systems. This project stands out by not requiring additional API keys, giving users complete command over their local and remote Macs directly through the Claude Desktop App. It offers a robust alternative to existing AI control solutions, specifically optimized for autonomous agents needing comprehensive desktop interaction capabilities without complex setups.

Installation

Getting started with MCP Remote macOS Use involves a few straightforward steps:

• Enable Screen Sharing on MacOs: Ensure screen sharing is active on your target Mac. (Skip if using a service like macstadium.com).
• Connect to your remote MacOs: Establish a connection to your remote Mac.
• Install Docker Desktop for local Mac: Docker Desktop is required on your local machine.
• Add this MCP server to Claude Desktop: Configure Claude Desktop to use the Docker image.

Example Claude Desktop configuration:

{
  "mcpServers": {
    "remote-macos-use": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "-e",
        "MACOS_USERNAME=your_macos_username",
        "-e",
        "MACOS_PASSWORD=your_macos_password",
        "-e",
        "MACOS_HOST=your_macos_hostname_or_ip",
        "--rm",
        "buryhuang/mcp-remote-macos-use:latest"
      ]
    }
  }
}

Examples

This project enables a wide range of autonomous AI applications on macOS, including:

• Research and Social Media: Researching Twitter and posting updates.
• Video Editing: Using applications like CapCut to create short highlight videos.
• AI Recruiter: Automating candidate information collection, qualifying applications, and sending screening sessions via Mail App.
• AI Marketing Intern: Engaging on LinkedIn and Twitter with automated following, liking, and commenting.

The server provides several tools for remote macOS control:

• remote_macos_get_screen: Capture a screenshot of the remote desktop.
• remote_macos_send_keys: Send keyboard input.
• remote_macos_mouse_move: Move the mouse cursor.
• remote_macos_mouse_click, remote_macos_mouse_double_click: Perform mouse clicks.
• remote_macos_mouse_scroll: Perform mouse scrolling.
• remote_macos_open_application: Open/activate an application.
• remote_macos_mouse_drag_n_drop: Perform drag and drop operations.

Why Use

MCP Remote macOS Use offers several compelling advantages for AI-driven macOS automation:

• No Extra API Costs: Leverage your existing Claude Pro plan for screen processing without additional API expenses.
• Minimal Setup: Only requires enabling Screen Sharing on the target Mac, with no extra software installation needed.
• Universal Compatibility: Designed to work across all macOS versions, ensuring future-proof integration.
• Native macOS Experience: Allows AI to operate within the rich macOS ecosystem, providing an unmatched user experience.
• Open Architecture: Supports universal LLM compatibility and model flexibility, integrating seamlessly with various LLM providers and the evolving MCP ecosystem.
• Effortless Deployment: Eliminates backend complexity, requiring zero setup on target machines beyond Screen Sharing.
• Streamlined Bootstrap Process: Utilizes Claude Desktop's polished UI for an intuitive user experience and instant productivity.
• WebRTC Support: Includes LiveKit integration for low-latency real-time screen sharing, improved performance, and better network efficiency.

Links

• GitHub Repository: https://github.com/baryhuang/mcp-remote-macos-use
• Docker Hub: https://hub.docker.com/r/buryhuang/mcp-remote-macos-use
• Model Context Protocol (MCP) Quickstart: https://modelcontextprotocol.io/quickstart/user