// the find
browser-use/web-ui
🖥️ Run AI Agent in your browser.
A Gradio-based web UI wrapper around the browser-use library, which lets AI agents control a browser to complete tasks. It's for people who want to drive browser-use without writing code — point it at a local URL, pick a model, type an instruction. Not a new agent framework; it's a front-end skin.
Supports a wide range of LLMs out of the box (OpenAI, Anthropic, Gemini, Ollama, DeepSeek) without needing code changes — just drop in your key and pick from the dropdown. The 'use your own browser' feature is genuinely useful: it piggybacks on your existing Chrome session so you don't have to re-auth into every site. Docker setup with a VNC viewer is a solid choice for running this headless — you can watch what the agent is actually doing without a local display. Persistent browser sessions between tasks mean multi-step workflows don't lose state.
This is almost entirely a thin UI over browser-use — if that library has a bug or a breaking change, this repo has no leverage; check the dependency before adopting. The test suite is skeletal: three test files that appear to exercise happy paths only, no mocking of Playwright interactions, nothing that would catch regressions in the agent loop. Last push was May 2026, changelog stops at January 2025 — maintenance cadence looks like it stalled after the initial hype. Gradio as the UI framework is fine for demos but it's an awkward choice for anything production-facing: no real auth, no multi-user isolation, session state is global per worker.