WebBrain
An open-source AI browser agent for Chrome, Firefox and Edge. Works offline with local models or connects to cloud APIs. No WebBrain account required.
WebBrain is an open-source AI browser agent for people who want help reading pages and completing web tasks in their existing browser. It runs in a sidebar in Chrome, Firefox and Chromium browsers; its Edge extension supports Windows only. The project uses the GPL-3.0-or-later license.
You can run it fully offline with a local LLM through llama.cpp, Ollama or LM Studio, or use an OpenAI-compatible server. It also connects to cloud services such as OpenAI, Claude and OpenRouter. With local models, page data stays on your machine. Cloud models receive the content they need to process. WebBrain requires no account and collects no telemetry.
Ask mode reads pages without changing them. It can summarize articles, answer questions about the current page and extract structured information from tables, forms and PDFs. Act mode handles clicks, form filling, uploads and downloads across multi-step tasks, with permission prompts before consequential actions. Successful runs can become reusable workflows, and scheduled tasks or page watches can trigger later work.
WebBrain can use separate text and vision models, so screenshot interpretation doesn't require the planning model to handle images. It trims conversation context and compresses screenshots to reduce model input. A local MCP bridge lets Claude Code, Codex, Cursor and other MCP clients delegate tasks to your signed-in Chromium tabs while retaining browser permission checks. The bridge doesn't support Firefox, whose browser automation also has fewer capabilities than Chrome's.
