Gemini Computer Use

This is a Playwright-based agent loop that wires up Gemini 2.5's Computer Use model to actually drive a browser. You give it a goal and a starting URL, it takes screenshots, asks the model what to do next, executes the actions, and repeats until the task is done or you hit the turn limit. The standout piece is the safety confirmation flow: risky actions get flagged and require human approval before execution. It defaults to Chromium but you can point it at Chrome, Edge, or Brave. The operational advice is solid: run this in a sandbox, use the exclude flag to block actions you don't trust, and keep the viewport at 1440x900. It's a clean reference implementation if you want to build browser automation on top of Gemini's multimodal function calling.

skills.sh installs: 1.2K
GitHub stars: 1K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add am-will/codex-skills --skill gemini-computer-use --agent claude-code

Installs into .claude/skills of the current project.

Files

Gemini Computer Use

Quick start

Source the env file and set your API key:

cp env.example env.sh
$EDITOR env.sh
source env.sh

Create a virtual environment and install dependencies:

python -m venv .venv
source .venv/bin/activate
pip install google-genai playwright
playwright install chromium

Run the agent script with a prompt:

python scripts/computer_use_agent.py \
  --prompt "Find the latest blog post title on example.com" \
  --start-url "https://example.com" \
  --turn-limit 6

Browser selection

Default: Playwright's bundled Chromium (no env vars required).
Choose a channel (Chrome/Edge) with COMPUTER_USE_BROWSER_CHANNEL.
Use a custom Chromium-based executable (e.g., Brave) with COMPUTER_USE_BROWSER_EXECUTABLE.

If both are set, COMPUTER_USE_BROWSER_EXECUTABLE takes precedence.