Gemini Computer Use Agent

•
2
•

Builds and runs Gemini 2.5 Computer Use browser automation agents using Playwright with built-in safety controls.

The Gemini Computer Use agent skill enables Claude to configure, run, and manage automated web browser workflows powered by the Gemini 2.5 Computer Use model and Playwright. It guides users through creating a continuous visual feedback loop—capturing viewport screenshots, parsing model function calls, executing UI interactions, and returning the updated browser state. The skill provides out-of-the-box support for multiple Chromium-based browser channels, customizable action exclusion filters, turn limits, and human-in-the-loop safety confirmation prompts for critical or risky UI operations.

Key Features

01Interactive safety confirmation prompts for high-risk UI actions
02Playwright integration supporting Chromium, Chrome, Edge, and custom executables
03Visual feedback pipeline passing screenshots and URLs back to the agent
042 GitHub stars
05Configurable turn limits and action blacklists to prevent unintended execution
06Autonomous browser control via Gemini 2.5 function-calling loops

Use Cases

01Automating multi-step web browsing tasks and dynamic form submissions
02Running automated end-to-end web testing with visual AI decision-making
03Navigating and extracting information from complex, JavaScript-heavy web applications