AI QA Engineer
An autonomous QA engineer you brief in plain words. It plans the cases, drives a real browser, gathers evidence and tells you what it found.
- Status
- In development
- Built with
- TypeScript, Playwright, Node.js
Where Kiwame runs inside your coding agent, this one runs on its own. You tell it what to test in plain words and it goes.
AI thinks, browser executes
That split is the whole design. The model plans and reasons; deterministic code does the clicking. Letting a model drive a browser click by click is slow, expensive and unrepeatable.
So one call plans the run, a rule engine decides pass or fail, and a vision model is invoked only when the rules can’t settle it. Every run prints the number of AI calls it made, because a test suite whose cost you can’t see is a test suite you stop running.
Working on a real app
Two things matter before it’s useful against something real. It needs a brief describing your product, which it keeps in a file you edit. And it needs to get past your login, which it does by borrowing a Chrome profile that is already signed in, rather than asking you to hand over credentials.
Any OpenAI-compatible provider works, so the model is yours to choose.