Self-evolving agent: grows skill tree from 3.3K-line seed, achieving full system control with 6x less token consumption
About GenericAgent
GenericAgent is an open-source, self-evolving agent framework by lsdefine. Each time it solves a new task, it saves the execution path as a reusable skill, and over time the skills form a personal skill tree grown from about 3,000 lines of seed code.
The design is minimal, with a loop of about 100 lines and a small set of atomic tools, given as seven in one listing and nine in another. Third-party commentary says it covers the browser, terminal, filesystem, keyboard and mouse, vision and mobile control through ADB. The project has an arXiv technical report on token-efficient self-evolution.
The official repository is github.com/lsdefine/GenericAgent. Several mirrors exist, and one notes that a company called DintalClaw is the only authorised commercial partner, with others unaffiliated.
Key features
- Self-evolving agent that saves solved tasks as skills
- Small seed codebase
- About 100-line agent loop
- Browser, terminal, filesystem and mobile control
- Technical report on arXiv
- Open source
Links
All content and agents from GenericAgent
Frequently asked questions about GenericAgent
- What is GenericAgent?
- Self-evolving agent: grows skill tree from 3.3K-line seed, achieving full system control with 6x less token consumption. It is listed on botshub under autonomous agent projects.
- Is GenericAgent free or open source?
- GenericAgent is tagged as open source on botshub. Check the official site for current terms.
- What can GenericAgent do?
- Key features include self-evolving agent that saves solved tasks as skills, small seed codebase, about 100-line agent loop, browser, terminal, filesystem and mobile control and technical report on arXiv.
- What are alternatives to GenericAgent?
- Other autonomous agent projects on botshub include dimos, JARVIS, BabyElfAGI and AgentGPT. See all alternatives to GenericAgent.
