Computer Use Preview
Reference browser agent driven by Gemini's computer use models
What it is
A Python command-line sample that lets a Gemini model operate a Chrome browser from a natural-language request, taking screenshots and issuing clicks and keystrokes. It runs the browser locally through Playwright or remotely on Browserbase, and can authenticate through the Gemini Developer API or through Vertex AI. Developers exploring Gemini's computer use capability can study it as a starting point for their own browser-controlling agents.