Chrome Bridge Automation

This connects Claude to your actual Chrome browser through the Midscene extension, so it can automate tasks while keeping all your logged-in sessions intact. It works entirely through screenshots and vision models (Gemini, Qwen, or similar), which means it can interact with anything visible on screen regardless of whether it's a React app, a Canvas game, or some ancient Flash-era interface. The critical constraint here is you have to run commands one at a time and wait for each to finish, because Claude needs to see the screenshot after each action before deciding what to do next. Commands can take a full minute since they involve AI inference. Best for workflows where preserving login state matters more than raw speed, like scraping data from internal tools or testing authenticated flows.

Install to Claude Code

npx -y skills add web-infra-dev/midscene-skills --skill chrome-bridge-automation --agent claude-code