Apps & data pipelines
jevqa is an open-source testing tool that uses Claude to write a checklist from a product spec and Jev to autonomously explore the web app for bugs.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
The maker reports detecting 22% of known WebTestBench bugs, while only about one in eight findings was confirmed. Sparse specifications lead to sparse tests, and complex flows and permissions remain weak spots. Page text goes to hosted Jev. Treat jevqa as an additional check rather than a QA replacement.
Start with a running web app and instructions explaining what it should do. The first run needs your access keys for Anthropic and TypeSafe to connect to their AI services. jevqa explores the screens and reports suspected bugs with screenshots. Check those findings yourself, and remember that page text is shared with hosted AI services.