eggbrid2/mobileClaw

Open Android AI agent runtime for phone control, app automation, VLM screen reading, skill routing, mini apps, and Mihomo VPN workflows.

What it solves

MobileClaw transforms an Android phone into a laboratory for AI agents, allowing users to pit different AI models and roles against each other in social games and competitive scenarios to test their reasoning, deception, and leadership capabilities.

How it works

It provides a game-style group room environment where users can assign different AI roles, each powered by a specific model and gateway. A system host manages the game flow by calling on speakers, collecting votes, handling hidden actions (such as "night" actions in traitor-hunt games), and announcing results to progress the game rounds.

Who it’s for

Developers and AI enthusiasts who want to conduct model-vs-model tests, run multi-agent simulations, or create custom elimination games on a mobile device.

Highlights

  • Multi-Model Arena: Support for different providers and models within a single group game.
  • Game-Specific Logic: Built-in flows for voting, elimination, and seat lockouts.
  • Privacy Controls: Secure handling of hidden actions to prevent accidental leaks of agent roles or moves.
  • Extended Agent Runtime: Includes capabilities for phone control, memory management, local/cloud model routing, and MiniAPP support.

Related

  • Project
    swarmclawai/swarmclaw

    A self-hosted AI agent runtime and multi-agent framework for creating autonomous agent swarms with durable memory, scheduling, and multi-provider LLM support.

  • Project
    ModalityDance/PalmClaw

    An AI agent framework that runs natively on Android, managing sessions and tool execution on-device while using external LLM APIs for inference.

  • Project
    aiming-lab/MetaClaw

    An agent proxy that enables AI assistants to meta-learn and evolve through real-world conversations using skill injection and asynchronous RL fine-tuning.

  • Project
    google-research/android_world

    An environment and benchmark for building and testing autonomous agents that can control Android devices across a wide variety of real-world apps.

  • Project
    Taewan-P/gpt_mobile

    A multi-provider AI chat client for Android that allows users to interact with various LLMs and deploy on-device agents with tool-calling capabilities.