GenericAgent
Minimal agent that controls your computer and saves solved tasks as skills
What it is
GenericAgent is a small Python agent framework, about 3K lines with nine basic tools, that gives a language model control over a local machine: browser, terminal, files, keyboard and mouse, screen reading, and Android phones over ADB. Each time it finishes a new task it records the steps as a reusable skill, so its abilities build up with use. It can run on models such as Claude, Gemini, Kimi, and MiniMax, among others, and keeps its context window small to cut token costs. Developers and tinkerers run it from a terminal or Streamlit UI to automate desktop, web, and phone chores.