Self-Evolving Autonomous Agent Framework

GenericAgent is a minimal, self-evolving autonomous agent framework designed to provide system-level control over local computers. With a core of approximately 3,000 lines of code, it utilizes 9 atomic tools and a compact Agent Loop of around 100 lines to enable various automation tasks such as browser control, terminal operations, and more.

Key features:

• Self-Evolving - Automatically crystallizes each task into a skill, allowing capabilities to grow with every use.
• Minimal Architecture - Comprises ~3K lines of core code with no complex dependencies.
• Strong Execution - Directly injects into real browsers while preserving login sessions.
• High Compatibility - Supports major models like Claude and Gemini, ensuring cross-platform functionality.
• Token Efficient - Operates with a context window of less than 30K tokens, significantly reducing resource consumption compared to other agents.

GenericAgent's unique self-evolution mechanism allows it to dynamically create new skills as it encounters new tasks, leading to a personalized skill tree that grows over time.