hkuds/autoagent

AutoAgent is a tool that lets you build and run AI agents without writing any code. Instead of programming, you describe what you want in plain conversational language, and the system figures out the rest. It was created by researchers at the Hong Kong University of Data Science and is available as an open-source project anyone can download and run. The core idea is that most AI agent frameworks still require developers. AutoAgent removes that barrier. You type a description of the kind of assistant or workflow you want, and the system constructs the underlying structure automatically. It can produce single agents for focused tasks or multi-agent pipelines where several specialized assistants collaborate on more complex problems. There are three ways to use it. The first is a ready-made research assistant mode that handles information gathering, analysis, and report writing. The second, called agent editor, is a conversation-based interface where you describe a custom agent and the system builds it step by step, including the tools it needs. The third, workflow editor, adds the ability to chain multiple agents together into a coordinated sequence. All three modes are accessible through a command-line interface after installation. Author comparisons in the README position the research assistant mode as competitive with commercial deep-research products at a fraction of the cost, since it runs on any major AI model, including open alternatives. You supply your own API keys for whichever AI provider you want to use. Installation involves Python and Docker. The README provides step-by-step setup instructions and guidance on connecting API keys. Community support is available via Slack and Discord. A related research paper on the underlying approach is also linked for readers who want more technical depth.