UI Verify

UI Verify is a hosted visual regression testing service that captures UI from Storybook, Playwright, Vitest, Bun or uploaded screenshots, diffs it against baselines on every pull request, and posts a check and comment on GitHub. An AI judge labels each change as an intended update, a regression or a flake, and an MCP server lets coding agents such as Claude Code, Cursor and Codex inspect and act on the results. The client packages (CLI, capture SDKs and agent skills) are open source under the MIT license.

What It Is

UI Verify plugs into CI and renders components and pages in the cloud. It compares them with accepted baselines and gates the pull request on the verdict, so reviewers see decisions rather than raw pixel diffs. Capture works through existing tests: Storybook stories, Playwright end-to-end tests, Vitest browser-mode tests, visual tests on bun test, or PNG screenshots from native, mobile and React Native harnesses.

How the Review Loop Works

When a PR opens, UI Verify screenshots the screens the change touches and comments with a summary. The AI judge reads the diff and gives a verdict with reasoning. Teammates can comment on exact pixels and accept, deny or ignore changes, and the decision returns to the PR. The docs state that nothing becomes a baseline on the judge's say-so, and that the check holds the merge until changes are accepted.

Agent Workflow Over MCP

A remote MCP server exposes tools to list projects and builds, get build and PR changesets, fetch per-story diffs and diff images, and accept, deny or ignore changes. A uiverify check command renders only the stories an agent is editing against the real CI baseline without posting a PR check or moving a baseline. Agent skills can be installed to set up visual testing and triage builds, and a Gemini CLI extension and Claude Code plugin are provided.

Capture and Determinism

The docs describe flake-resistant capture that pins animations, timers, fonts and async data, automatic flake detection by re-rendering, and skipping unchanged stories while carrying their results forward. Modes cover Chromium, Firefox and WebKit at desktop and phone widths, plus light and dark themes, with versioned baselines and Figma design baselines. GitHub is the supported integration; GitLab and Bitbucket are listed as roadmap items.