UI Verify

Documentation

Everything you need to set up visual regression testing with UI Verify, understand how it works, and keep your captures fast, cheap, and deterministic. New here? Start with what visual testing is.

Getting started

What is visual regression testing?

The concept, in plain terms: render the UI, screenshot it, diff it against a baseline, review what moved.

Read

Storybook visual regression testing

Your existing stories are the tests. Build Storybook, upload it, get a diff check on every PR.

Read

Playwright visual regression testing

No Storybook needed. Archive the real pages your Playwright tests drive, then diff them every PR.

Read

Core concepts

How visual regression testing works

The pipeline end to end: upload, render, diff, AI triage, review - one CI step, no browsers to manage.

Read

Baselines and branches

What a baseline is, how it is chosen per branch, and the one rule for when it updates: only on accept.

Read

Skip unchanged stories with --only-changed

Most PRs touch a few components. Render those, carry the rest forward - same coverage, a fraction of the work.

Read

Automated visual review with an AI judge

A diff tool tells you that pixels changed. The AI judge tells you whether the change was the point of the PR.

Read

UI Verify vs Chromatic

The same screenshot diffing you know from Chromatic, plus an AI judge and agent review - and it works without Storybook.

Read

Recipes

Reduce visual test snapshots and cost

The story count is the cost. Two moves collapse an exploding variant matrix into a handful of dense stories.

Read

Fix flaky visual tests

A flaky diff is run-to-run variation the screenshot captured faithfully. Kill the variation at record time.

Read

Triage visual changes from your coding agent

If an agent wrote the PR, an agent can review the pixels. List changes, read the verdict, accept - over MCP.

Read

Troubleshooting UI Verify

The setup problems that fail silently - no PR check, an empty archive, phantom diffs - and the one-line fix for each.

Read