Analyze spreadsheets or databases by chatting with the Data Agent, which writes and runs the code on your behalf.
Let the Plugins Agent automatically pick from 200+ external tools to answer questions about weather, shopping, or scientific calculations.
Automate web tasks by asking the Web Agent to navigate websites, fill out forms, or retrieve information from any page.
Build a custom AI assistant platform by swapping in different language models or adding new agent types to the open codebase.
Requires Docker and multiple API keys, each component needs its own environment file configured before starting containers.
OpenAgents is an open-source platform from a university research lab that lets you run AI assistants capable of doing real tasks through a chat interface. Think of it as a self-hosted version of ChatGPT's advanced features, but with code you can read, modify, and deploy yourself. The project comes with a web UI and a backend server, so you can set it up on your own computer or server without relying on a third-party service. The platform ships with three ready-to-use agents. The first is a Data Agent that can write and run Python or SQL code to analyze spreadsheets, databases, and files, then produce charts or summaries from the results. The second is a Plugins Agent that connects to over 200 external tools covering shopping, weather lookups, scientific calculations, and much more, choosing the right tools automatically based on what you ask. The third is a Web Agent that controls a real Chrome browser to navigate websites, fill forms, and retrieve information on your behalf. The intended audience is split between everyday users and developers. Regular users get a chat interface designed to handle common failures gracefully and respond quickly. Developers and researchers get a fully open codebase they can extend with new agent types, swap in different underlying language models, or use as a starting point for studying how these systems behave in practice. The paper accompanying the project, published on arXiv, documents both the design decisions and the open challenges the authors encountered when deploying agents to real users. Setting up a local copy requires Docker and a few API keys. The README walks through cloning the repository, configuring environment files for each component, and starting the frontend and backend containers. An online demo was hosted by the lab for free use, though the README notes that high traffic sometimes affects response times. The project is licensed under Apache 2.0.
← xlang-ai on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.