Give it one goal in plain English and it drives a real browser to get there. On each page it lists the buttons and fields it can see, and Jev chooses what to do next and which element to do it on.
A small language model is called only when something has to be typed, like a city name. The team behind Browser Use reports a Google Flights search from Zurich to London finishing in about seven seconds.
Needs a TypeSafe API key and a key for the small text model.
What you get
- Full Python source code, MIT licensed
- The agent loop, examples and tests from the repository
- Setup steps using uv
Jev
Python
FastAPI
MCP
Hermes
CodeRabbit participant
