How to set up visual regression testing

You do not have to write new tests to start. UI Verify reuses whatever already renders your components or pages, screenshots each state, diffs it against a baseline, and posts a check on every pull request. The one decision that shapes your setup is where those screenshots come from, so start there.

Pick the capture method that matches what your project already has. Every path lands in the same place - a diff check on each pull request - so this choice is about your existing tooling, not features you gain or lose.

Whichever capture method you pick, the moving parts are the same four steps.

1. Get a project API key from uiverify.ai and store it in CI as UIVERIFY_API_KEY.
2. Run the uiverify CLI to upload a build. Your first run captures every state as new and flags it for review; accept those to set your baselines.
3. On later runs, UI Verify diffs each capture against its baseline and reports what moved.
4. The GitHub App posts a check and a comment on each pull request, where you or an AI judge accept or reject each change.

Each setup guide opens with a copy-paste prompt for your coding agent. Paste it into Claude Code, Cursor, or any agent and it reads the guide, installs the packages, adds the config, and uploads the first build for you. It is the same work the guides describe by hand, done in one step, which is why the quickest start is to open the guide for your capture method and hand its prompt to your agent.

Choose by what your project already renders. If you have Storybook, use the Storybook guide, since each story is already a capture target. With no Storybook, use Vitest for component tests or Playwright for real pages and flows. For screenshots produced by any other tool, including native and mobile, use the custom screenshots guide.

You do not need existing tests or stories to start. Your coding agent writes them: point it at the components or pages worth covering and it scaffolds the Vitest component tests, Storybook stories, or Playwright tests, then UI Verify captures each one. Use Vitest or Storybook for components and Playwright for an app's pages and flows. Then add a rule to your agent's instructions, a CLAUDE.md rule and a lint check, that asks for a test on every new component or page, so coverage grows as the agent builds. See Stop writing Storybook stories by hand.

No. UI Verify captures the components and pages your existing tests or stories already render, so there is nothing new to author. See Visual testing setup: no new tests to write.

No. Storybook is one capture method, not a requirement. Vitest browser-mode tests, Playwright tests, and custom screenshots all capture without it, so a project with no Storybook sets up just as fast.