Catch bad
design before
it ships.
Crit opens the real rendered screen of every pull request, judges it against your team’s own standard, and flags anything off-brand, right where code gets approved.
Add invoice history table
AI-likelihood medium · parity 88%
Table text 13px, under 16px floor
Action uses indigo #4f46e5, off-palette
Design review that doesn’t wait on someone noticing.
Most off-brand screens ship not because nobody cares, but because nobody opened the preview in time. Crit removes the human bottleneck.
- Off-brand colour and spacing slip through to production
- AI-generated screens ship looking templated and generic
- Feedback waits on a busy reviewer opening the preview build
- Contrast bugs hide in the theme nobody screenshotted
- Every PR gets an automatic design verdict, on the real screen
- The AI-generated look is flagged before it reaches users
- Feedback arrives in seconds, with one-click fixes attached
- Both light and dark themes reviewed, every single time
What a linter can’t see, Crit does.
The AI-slop tell.
Templated hero-and-three-cards, weak hierarchy, default shadows, indigo-on-white, the generic machine-made look Crit is purpose-built to detect and score.
On-brand colour
Every fill, border and state checked against your real palette.
Type & spacing
Sizes, weights and rhythm measured on the rendered screen.
Pull request in. Verdict out. No dashboards to babysit.
Install on a repo
Point Crit at a repository. It reads your Tailwind config, tokens and components to learn your Standard, read-only, two minutes.
Open a pull request
When a PR touches the UI, Crit finds your preview deploy, opens it in a real browser, and captures the rendered screen.
Get a verdict in the PR
A score, an AI-likelihood read, annotated shots and one-click fixes, as a comment and a status check that can block merge.
Spot what a linter can’t.
Crit looks at the real pixels, not the code. Tap a marker to see what it catches, and the one-click fix it hands your team.
Table text is 13px, under your 16px floor and fails AA contrast on the row hover.
Regression sees what changed. Crit sees whether it’s any good.
| Visual regression | Code review | Crit | |
|---|---|---|---|
| Notices a screen changed | |||
| Catches logic & code bugs | |||
| Judges visual quality & taste | |||
| Knows your brand & standard | |||
| Flags the AI-generated look | |||
| Can block the merge |
A complete design review, where you already work.
Beautiful in both themes.
Every screen ships in light and dark. Crit reviews both, so contrast bugs never hide in the mode you didn’t screenshot.
One-click fixes
Every issue ships with a pasteable prompt for Cursor, Claude Code or v0.
Trust by default
Read-only, least-privilege GitHub access. We never train on your code.
Judged on your scale
Spacing and type measured against your own tokens, not a generic rulebook.
“Crit caught three off-brand screens in a week that our whole team had scrolled past in review. It’s the design eye we never had time to be.”
Maya Adeyemi
Head of Design · Lattice
The trust questions, answered.
Connecting an outside service to private code is a big deal. Here’s exactly how Crit handles it.
Read the full security pageMake design part of the merge.
Connect a repo in two minutes. Crit reviews your next pull request automatically.