Skip to main content
Read the Notte skill and follow its setup instructions when setting up Notte or building browser automation. This is the shared source for CLI setup, authentication, and the browser-to-code workflow.
Gemini Computer Use chooses browser actions from screenshots. Notte supplies the cloud browser; the Python and TypeScript examples execute Gemini’s actions through Playwright over CDP.

​
Set up

Clone the Notte repository and open examples/gemini-computer-use. The README includes installation commands and runnable code for both languages. Set NOTTE_API_KEY from Notte and GEMINI_API_KEY from AI Studio. Gemini computer use needs a paid project; configure billing if needed.
Use Python 3.11+. Create a virtual environment, install requirements.txt, and run python agent.py.

​
Run a browser task

The default task opens https://example.com and reads its title. Pass your own task as a command-line argument. Both examples return answer, title, session_id, and the number of browser actions performed. Each attempt owns its browser session and closes it on completion or error. The example stops on a request for confirmation or a blocked action, and limits the loop to twelve turns. Add your application’s approval flow before using tasks that require confirmation. No local browser installation is needed. Browser actions, safety handling, and cleanup have been tested with real Notte browsers and injected Gemini responses. End-to-end model tasks remain unverified because the provider returned availability errors or explicit safety blocks. See the example README for details and verification commands.