openagent

OpenAgent is an open-source, self-hosted AI assistant that you can run on your own computer or server. It connects to over 30 AI model providers, including OpenAI, Anthropic, Google, and many others, so you can switch between models without changing any settings in your code. The whole application ships as a single file you download and run, with no complicated installation steps. You open a browser, go to a local address, and start chatting.

One of its main features is a knowledge base you build yourself. You upload your own documents, like PDFs or Word files, and the system breaks them into searchable chunks. When you ask a question, it pulls the most relevant pieces from your documents before sending anything to the AI, so the responses are based on your actual files rather than just the model's training data.

OpenAgent can also take actions on your behalf through what the project calls agent loops. It can open a real browser and navigate websites, click buttons, fill out forms, and take screenshots. It can search the web, run shell commands, read and write Office files, and connect to any compatible external tool through a standard called MCP. Every step the agent takes is shown to you so you can follow along.

For more complex tasks, there is a visual workflow builder where you draw out multi-step processes using a drag-and-drop interface. You can add conditions, run steps in parallel, and schedule workflows to run on a timer. The platform also tracks how many tokens each model uses so you can monitor costs.

The project supports Windows, macOS, and Linux, including native Windows without needing any extra compatibility layers. It has built-in user authentication via standard protocols and supports multiple users with separate workspaces. A full API with documentation is available so you can access all features programmatically.