Seatbelt logoSeatbeltGet Seatbelt

QA your AI-coded app.

Seatbelt maps the critical flows in your app, replays them in a real browser, and gives you the verdict, the evidence, and an LLM fix prompt.

Get Seatbelt See how it works
  • No signup
  • No credit card
  • Two commands to start
localhost:3000/checkout
Critical flows3/5
  • loginpassed
  • onboardingpassed
  • uploadpassed
  • checkoutfailed
  • dashboardblocked
DO NOT MERGE

checkout · failed at step 6

expected success redirect → got entitlement locked

screenshotvideotraceconsolenetwork

Investigate checkout success: the entitlement isn't unlocking after payment. Failing step 6, redirect never fires…

Demo video coming: red failure → fix prompt → green rerun

Built for people shipping AI-generated code with Claude, Cursor, or Codex.

The AI coding hangover

Code generation got 10× faster. Checking the result didn’t. That part still lands on you — after every AI change.

1

The diff landed

Claude, Cursor, or Codex just rewrote 20 files. It looks fine. It compiles.

2

Now click it all. By hand.

Login, onboarding, upload, checkout, entitlements, dashboard — one more time.

3

So you skip it

It’s draining after a long session. The one check you skip is the one that breaks prod.

4

Unit tests miss it

They check pieces. AI-coded apps break where the pieces are wired together.

How Seatbelt works

From install to completed browser testing in four steps. No test suite to babysit.

terminal
$ npm install getseatbelt$ npx seatbelt init✓ Seatbelt is set up and your first browser QA is ready

Two commands. Real browser QA.

Add Seatbelt, let it map your flows, then run the browser check before you merge.

  • Works in your repo with two commands.
  • It detects your user flows and runs a live browser check.
  • No signup, no credit card.
$ npm install getseatbelt$ npx seatbelt init

You watch the browser do it.

Seatbelt drives a real browser through your flows, so the result is something you can actually trust.

localhost:3000/checkout
Critical flows3/5
  • loginpassed
  • onboardingpassed
  • uploadpassed
  • checkoutfailed
  • dashboardblocked
DO NOT MERGE

checkout · failed at step 6

expected success redirect → got entitlement locked

screenshotvideotraceconsolenetwork

Investigate checkout success: the entitlement isn't unlocking after payment. Failing step 6, redirect never fires…

Visible replay

A real browser window — not a green check you have to take on faith.

Fresh test state

A clean test user each run, so a pass is real — not a leftover session.

Evidence, saved

Video, screenshots, and logs saved locally on every run.

A report you can act on.

After each run, Seatbelt gives you the verdict, the failed step if something broke, the evidence, and an LLM fix prompt you can send to your coding agent.

Every run captures

  • Merge recommendation
  • Pass / fail by flow
  • The exact failed step
  • Expected vs actual
  • Screenshots
  • Video / replay
  • Console errors
  • Network failures
  • Repro steps
  • Suspected cause
DO NOT MERGE

checkout failed — expected a success redirect, got an entitlement still locked after payment.

Investigate the checkout success flow. After a completed test-mode payment, the entitlement never unlocks and the success redirect never fires (failed at step 6). Check the webhook handler and the entitlement write. Repro + trace attached.

Questions, answered

Is this just test generation?

No. Seatbelt manages the QA loop around your app: it maps critical flows, keeps them fresh as your code changes, replays them in a real browser, captures evidence, and gives you the fix prompt. You don’t babysit a test suite to know if checkout broke.

Does it open a real browser?

Yes. You watch it click, type, upload files, complete checkout, or run the flows that matter in your app. Seeing the run is what makes Seatbelt trustworthy.

Do I need my own OpenAI key?

No. Seatbelt’s AI features run through Seatbelt. You don’t bring or paste an OpenAI key.

Do I need to paste an API token?

No. Normal use opens your browser and connects your account automatically. Tokens are only for advanced CI/dev setups.

Can I use it before every merge or commit?

Yes — run it any time you’re unsure about your app, not just before a PR. After an AI diff, run seatbelt run, watch the browser replay, then commit, merge, or send the fix prompt back to Claude, Cursor, or Codex.

How do I reset test accounts?

During seatbelt init, Seatbelt can generate a reset prompt for your coding agent. It adds a safe reset script to your app and wires it into your Seatbelt config, so every run starts from a clean test user. Say yes during init and you won’t need to run setup-reset for that repo.

Does it run locally?

Seatbelt runs from your repo and opens a browser against your app. It’s built for local dev and pre-merge checks.

Does it replace unit tests?

No. Unit tests check pieces. Seatbelt checks the user journeys where those pieces are wired together: login, onboarding, upload, checkout, entitlements, dashboards, and the flows specific to your app.

Start free with real browser QA.

Run Seatbelt in your repo with no login and no credit card. Upgrade when Seatbelt becomes part of your routine.

Get Seatbelt View pricing