AlexGladkov/claude-in-mobile

What it solves

It allows AI assistants (like Claude) to interact with and control mobile devices and other platforms, enabling them to perform actions such as taking screenshots, interacting with apps, and debugging live applications.

How it works

The project provides a Model Context Protocol (MCP) server that connects an AI client to various platforms via plugins. It supports Android, iOS, Web, Desktop, and Aurora. Depending on the platform, it uses underlying tools like adb for Android or xcrun for iOS to execute commands, taps, swipes, and typing.

Who it’s for

Developers and testers who want to give their AI agents the ability to drive apps, perform visual regression testing, accessibility audits, and runtime debugging across different operating systems.

Highlights

• Multi-platform support: Works with Android, iOS, Web, Desktop, and Aurora.
• Device Control: Can perform taps, swipes, typing, and app launches.
• Deep Debugging: Supports breakpoints, stepping, and variable inspection via Android JDWP and iOS LLDB.
• Testing Tools: Includes capabilities for visual regression, performance monitoring, and sensor/network simulation.
• Modular Architecture: Available as a slim base (mcp-devices) with optional plugins or an all-in-one bundle (claude-in-mobile).