๐ Overview
GenericAgent is a revolutionary, self-evolving autonomous agent framework built on a minimalist core of approximately 3,000 lines of code. It empowers any Large Language Model (LLM) with comprehensive system-level control over local computers and mobile devices through just 9 atomic tools and a lean 100-line Agent Loop. Its groundbreaking design philosophy focuses on "don't preload skills โ evolve them," allowing the agent to grow its capabilities organically over time.
The core innovation lies in its self-evolution mechanism: every time GenericAgent successfully completes a new task, it automatically crystallizes the execution path into a reusable skill. This process builds a personalized, ever-expanding skill tree, tailored to your unique usage patterns, all originating from a compact seed of code.
Self-Bootstrap Proof: A remarkable testament to its autonomy, GenericAgent has completed all repository tasks, from initial Git setup to commit messages, entirely on its own โ the author never manually opened a terminal.
๐ Key Features
- Autonomous Skill Evolution: Automatically transforms task solutions into reusable skills, continuously expanding the agent's capabilities and forming a personalized skill tree.
- Minimalist Architecture: A remarkably small codebase (~3K lines core, ~100 lines Agent Loop) with zero complex dependencies, ensuring effortless deployment and minimal overhead.
- Robust System Control: Integrates directly with real browsers (preserving login sessions) and leverages 9 atomic tools for deep control over terminals, filesystems, keyboard/mouse input, screen vision, and mobile devices (via ADB).
- Broad Model Compatibility: Seamlessly supports major LLMs including Claude, Gemini, Kimi, and MiniMax, offering cross-platform versatility.
- Cost-Efficient Operation: Achieves full system control with significantly reduced token consumption (6x less).
๐ Typical Use Cases
GenericAgent's self-evolving nature unlocks powerful applications by learning and adapting to your specific needs.
- Automated Task Execution: Handle repetitive tasks like ordering food, tracking expenses, or sending batch messages with unparalleled autonomy.
- Advanced Web Automation: Navigate complex websites, perform quantitative stock screenings, or continuously monitor and summarize web content without explicit programming.
- System Management & Development: From installing dependencies and writing scripts to managing files and orchestrating development workflows, GenericAgent learns and adapts to streamline system operations.
- Personalized Digital Assistant: Develop a truly bespoke assistant that understands and executes tasks based on your unique workflows and accumulated experience.
Example Evolution:
| User Request | First-Time Execution | Subsequent Execution |
|---|---|---|
| "Monitor stocks..." | Installs tools, builds screening logic, sets up cron. | "One-line start." |
| "Send file via Gmail" | Configures OAuth, writes sending script. | "Ready to use." |
| "Read WeChat..." | Installs deps, reverse-engineers DB, writes script. | "One-line invoke." |
The longer you use GenericAgent, the more sophisticated your personalized skill tree becomes, offering efficient, one-line command execution for tasks it has previously learned.
