GenericAgent is a minimalist, self-evolving autonomous agent framework designed to grant LLMs system-level control over local computers, enabling them to autonomously perform tasks and evolve their capabilities over time.
Source: per README View on GitHub →GenericAgent is gaining attention due to its minimalist design, self-evolving capabilities, and the ability to achieve full system control with significantly less token consumption compared to other agents. Its unique approach to skill crystallization and layered memory system stands out in the field of autonomous agents.
Source: Synthesis of README and project traitsAutomatically crystallizes each task into a skill, forming a personal skill tree that grows with usage. This feature is implemented through autonomous exploration, execution path crystallization, and skill writing to memory layers.
Source: per READMEThe core codebase is approximately 3K lines, with an Agent Loop of ~100 lines, ensuring minimal deployment overhead and no complex dependencies.
Source: per READMEInjects into a real browser to preserve login sessions and uses 9 atomic tools to take direct control of the system, ensuring robust execution.
Source: per READMESupports major LLM models like Claude, Gemini, Kimi, and MiniMax, and is cross-platform.
Source: per READMEConsumes less than 30K context window, with layered memory ensuring the right knowledge is always in scope, leading to higher success rates at a lower cost.
Source: per READMEThe architecture is inferred to be based on a layered memory system, an autonomous execution loop, and a minimal toolset. The layered memory system includes meta rules, insight index, global facts, task skills/SOPs, and session archive. The autonomous execution loop perceives the environment, reasons about tasks, executes tools, writes experience to memory, and loops back. The toolset is minimal, with 9 atomic tools.
Source: Code tree + dependency filesCenter: project; inner ring: core feature modules; outer ring: key dependencies. Auto-generated from core_features and tech_stack.key_deps.
requestsbeautifulsoup4bottlesimple-websocket-serverGenericAgent is suitable for developers and technical decision-makers looking to automate complex tasks, such as web browsing, stock monitoring, expense tracking, and batch messaging. It is also useful for creating custom autonomous agents for various applications.
Source: READMENo release records available.
Source: GitHub ReleasesGenericAgent is a promising open-source project for those interested in creating autonomous agents with minimal code and high efficiency. Its unique approach to self-evolution and skill crystallization makes it a valuable tool for developers and technical decision-makers looking to automate complex tasks and explore the capabilities of LLMs.