Panda is an on-device AI agent for Android that automates complex, multi-step tasks by understanding natural language commands and interacting with the phone's UI. It is designed for users who want a more accessible and personalized mobile experience, acting as a proactive operator to handle tasks across various applications.

How It Works

Panda utilizes a multi-agent system architecture written in Kotlin. Its "Eyes & Hands" component, an Android Accessibility Service, enables it to perceive the screen's UI elements and execute touch gestures. The "Brain" is powered by LLM models for reasoning and planning, while the "Agent: Operator" acts as the executor, managing tasks with a "Notepad" for memory. This separation of concerns allows for complex and reliable operation.

Quick Start & Requirements

• Installation: Clone the repository, set up API keys in local.properties (including sdk.dir and GEMINI_API_KEYS), and open the project in Android Studio.
• Prerequisites: Android Studio, an Android device or emulator (API level 26+), and Gemini API keys.
• Setup: Enable the "Panda" Accessibility Service within the app's settings upon first run.
• Docs/Demos: Demos are available via provided GitHub asset links.

Highlighted Details

• Proactive, on-device AI agent for Android.
• Intelligent UI automation using Accessibility Service.
• Persistent, personalized local memory for learned procedures and user preferences.
• High-quality voice output via GCS Chirp.

Maintenance & Community

Contributions are welcome via issues or pull requests.

Licensing & Compatibility

Licensed under the MIT License, generally permissive for commercial use and closed-source linking.

Limitations & Caveats

The project is described as a "powerful proof-of-concept," and its roadmap indicates it is still under active development. The "What's Next" section notes that the roadmap list is not updated.