Hi Product Hunt, I'm the maintainer of Waterfall AI Test. I've spent years maintaining Playwright test suites, and the tiring part was never the code — it was the loop of reading requirements, guessing what deserves coverage, and then losing track of what was tried when the UI changed. Chat-based AI helpers made this worse, not better: everything lived in a scrolling conversation you couldn't audit. So I built an open-source visual workbench for Playwright test agents. It takes a requirement document and walks it through stages you can see and edit: test modules, a test plan, generated scripts, real browser runs, and a verified test set with reports, videos, and traces. Nothing enters the test set until it has actually passed. Failed runs keep their evidence, get repaired, and re-run. Everything is versioned in git, so AI-generated tests become a maintainable team asset instead of a one-off answer. It's built for QA engineers and developers who own automation — especially small teams without a dedicated automation person. Honest beta note: runs cost API money, and gnarly flows still need a human. You can try the live demo at http://43.130.174.131:5000/ (demouser / demouser) — the code is open source: https://github.com/jiongfeng/wat... I'd love feedback on where the human-in-the-loop steps feel like too much friction. Tell me what breaks your workflow — I'm around all day to answer.

Waterfall AI Test

Turn requirements into verified Playwright tests

Turn requirements into verified Playwright tests