Your agent can catch website visual regressions

Screenshots a set of key URLs every morning, vision-compares each against the baseline image saved on your drive, and on meaningful visual drift publishes a side-by-side diff board to clor.app behind basic auth and emails you. It ignores trivial rendering noise and only flags real layout or content regressions, and once a change is acknowledged on a clean run it adopts the new screenshot as the baseline. Does nothing on a run where every page matches its baseline within tolerance.