Mobile-Agent is a family of autonomous agents designed for complex task automation on mobile devices and PCs. It targets researchers and developers building AI assistants capable of understanding and interacting with graphical user interfaces, offering a multi-agent collaboration framework for enhanced navigation and task completion.

How It Works

The system employs a multi-agent collaboration approach, where specialized agents work together to achieve complex goals. Mobile-Agent-v2, for instance, uses effective navigation strategies via multi-agent collaboration for mobile device operations. Mobile-Agent-E introduces a hierarchical multi-agent framework with self-evolution capabilities based on past experiences, aiming for stronger performance on intricate, long-horizon tasks. PC-Agent extends this to PC operations on Mac and Windows.

Quick Start & Requirements

• Demo: Available on Hugging Face Space and ModelScope.
• Dependencies: Specific requirements vary by version; v3 is noted to use open-source models with 8GB memory and 10-15s reasoning per operation.

Highlighted Details

• Mobile-Agent-v2 accepted to NeurIPS 2024.
• Mobile-Agent won Best Demo Award at CCL 2024.
• PC-Agent supports Mac and Windows platforms.
• Mobile-Agent-E features self-evolution for complex tasks.

Maintenance & Community

The project has multiple associated papers and citations, indicating active research and development. Links to demos and related projects are provided.

Licensing & Compatibility

The README does not explicitly state the license type or any restrictions for commercial use or closed-source linking.

Limitations & Caveats

The project is presented as a research initiative with multiple evolving versions (v2, v3, PC-Agent, Mobile-Agent-E). Specific performance metrics and stability for production use are not detailed.