Automated visual review with an AI judge

A diff tool tells you *that* pixels changed. It cannot tell you *whether* the change was the point of the pull request or a mistake. On a busy repo that leaves you a review queue full of intended changes to click through one by one. UI Verify runs an AI judge over each changed story to do that triage first.

For each changed story the judge looks at the baseline, the candidate, the regions that changed, and what the pull request says it does (its title and description), and produces a verdict: a likely regression, or a likely intended change, with a short reason. Likely regressions sort to the top of the review queue and likely intended changes to the bottom, so the changes worth a close look come first.

No. The judge never changes a baseline on its own. It ranks and explains; promoting a candidate to a baseline is always an explicit action, taken by a human or by an agent you authorized. The pull-request check does not read the verdict either: it holds the PR until each change is reviewed, and a denied change fails it. A green build never means the AI quietly accepted something for you.

Yes. UI Verify checks a change is real by re-rendering the story - a diff that does not reproduce is auto-ignored as a flake and collapses out of the queue, so your review only spends time on changes that actually happened. See automatic flake detection.

Yes. Because the diffs and verdicts are exposed over MCP, your coding agent can pull them straight into the pull-request flow - list the changes, read the judge's reasoning, accept the intended ones - without you leaving the terminal. See Triage with your agent.