Point it at a URL. Get a triaged report with screenshots and video. Re-run after every fix to verify it's gone.
Real Chromium browser. The agent figures out what kind of site it is, picks the flows worth testing, and tries them — clicking, typing, navigating, edge cases included. No fixtures, no page objects, no recorders.
An LLM watches the agent explore your site and writes a structured test plan — what to test, with what data, in what order. The plan is stable across runs, so you can compare results over time without writing a single line of test code.
After every fix, hit Re-run. The same plan executes again and a diff card tells you what's new, fixed, regressed, or still broken — keyed by stable bug fingerprint. The full bug-finding loop, not just detection.