Let your coding agent click through your app in Chrome and fix what breaks
Idea + starter prompt, not run by us yet
For: Developers testing web features on their own computer
Once the Chrome extension is connected, Claude Code uses the developer's real Chrome session, and each new site needs a one time OK. Asked for a table filter, it wrote the component, opened the page, tried it, saw the URL did not update, fixed the routing and confirmed after a reload, in three rounds. A visible Chrome window is required, so it cannot run in CI or headless.
Steps
- Connect the Chrome extension to Claude Code on your own computer.
- Use a test account, since it shares your real browser logins.
- Ask it to build the feature, open the page, try it and read console errors.
- Check the fix yourself before you commit.
- Watch token use, since browser steps use a lot of context.
Starter prompt
Add [feature] to [page] in my app running at [localhost URL]. After you write the code, open the page in Chrome, try the feature like a user would, and read the console for errors. Fix what you find and test again. Use only my local site and test account [test account name]. Do not open other sites or submit real data. Tell me what you changed and show me the diff before I commit.
Original post: test-lab.ai · found via agentusecases.ai. Written up in our own words. We haven't run it, so test it on made-up data first.