Hi HN,
I built an open-source CLI that records a user flow in the browser and generates UI test cases from it.
It captures the interaction as a Playwright script and then generates additional test cases (including variations like invalid inputs, edge cases, etc.).
The goal is to help teams get initial UI test coverage faster. Writing tests from scratch is time-consuming, so this tries to reduce that initial effort.
The tool is fully open source and runs locally. You bring your own LLM key, and nothing is sent anywhere except your chosen provider.
Would really appreciate feedback on:
accuracy of generated tests
where it’s useful
where it breaks
The part about newer endpoints having significantly higher auth failure rates felt very real. In most teams I’ve seen, older endpoints get hardened over time but new ones ship fast and testing doesn’t catch up. This report kind of quantifies that gap.