Gemini Computer Use: Automate a Browser in Python

Gemini Can Now Drive Your Browser

On June 25, 2026 Google flipped a switch that a lot of agent builders had been waiting for: computer_use is now a built-in tool inside Gemini 3.5 Flash, callable straight from the Gemini API. No separate preview model, no special access tier. You hand the model a screenshot and a goal, and it hands back a concrete UI action: click at (x, y), type this text, scroll down, navigate to this URL. Your code runs the action in a real browser, screenshots the result, and the loop repeats until the task is done.