MaaMCP provides an MCP (Model Context Protocol) server built upon the MaaFramework, enabling AI assistants to automate Android devices and Windows desktops. It offers AI agents the capability to interact with applications, perform screen recognition, execute precise UI actions, and manage multiple devices concurrently. The project's core benefit is bridging AI intelligence with direct device control, facilitating complex automation tasks and generating reusable automation scripts through its unique Pipeline generation feature.

How It Works

MaaMCP acts as a standardized interface, translating AI commands into actions executable on Android (via ADB) and Windows. It leverages MaaFramework's capabilities for device discovery, connection, screen capture, OCR-based recognition, and control operations like clicking, swiping, and text input. A key innovation is its Pipeline generation system, where AI-driven actions are automatically converted into a JSON format, allowing for repeatable and robust automation workflows.

Quick Start & Requirements

• Installation: Recommended: uvx maa-mcp. Alternatives: pip install maa-mcp or from source (git clone https://github.com/MistEO/MaaMCP.git, pip install -e .).
• Prerequisites: ADB for Android automation. Python environment.
• Links: MaaFramework: https://github.com/MaaXYZ/MaaFramework. PyPI: https://pypi.org/project/maa-mcp/. A Bilibili video demo is mentioned but not directly linked.

Highlighted Details

• Unified automation for Android (ADB) and Windows desktops through a standardized MCP interface for AI assistants.
• Background operation for Windows automation, allowing users to continue working without mouse/keyboard interference.
• Support for multi-device and multi-window collaboration, enabling simultaneous control over several targets.
• AI-driven Pipeline generation: Translates executed automation sequences into reusable JSON scripts, complete with validation and self-healing capabilities.
• Intelligent screen recognition using OCR for precise UI element identification and interaction.

Maintenance & Community

The provided README does not detail specific contributors, sponsorships, or community channels like Discord or Slack.

Licensing & Compatibility

Licensed under GNU AGPL v3. This copyleft license requires derivative works to be distributed under the same license, potentially impacting commercial use or integration into closed-source projects.

Limitations & Caveats

Windows automation may be hindered by anti-cheat mechanisms in games or certain applications. MaaMCP requires administrator privileges if the target application runs with them. Minimized windows are not supported. If background screenshotting or input fails, the AI might resort to foreground control, occupying the mouse and keyboard. OCR model downloads may occasionally fail, necessitating manual checks or retries.