AI QA demo
A real run. No signup.
mini caught a cart bug in our own demo app. No customer data.
- Start with the suite mini wrote.12 tests, each a scenario plus the acceptance criteria that must hold.
- See what it flagged.4 of 12 passed. The cart says $85.48; its lines add to $82.48.
- Then it’s your call.A recording and a fix prompt. Not a bug? Say why, once.
Hang on. That’s not $82.48.Cart Management
Manage Cart Contents
Critical7/8 passed · Android
- Expected
- Confirm the price summary shows a subtotal, delivery fee, and grand total that reflect all items
- Actual
- Cart subtotal does not match the sum of line prices — overstates by $1.50 per cart line, inflating the grand total.
What we need to run it on your app.
A build
Android .apk (x86-64) or .apks, an iOS simulator build, or your repo.
TestFlight, for physical iPhones on request
Switched on per workspace. mini installs your app from TestFlight.
Test accounts
Staging sign-ins and test bypasses, like a fixed OTP. Never production credentials.
A staging URL, for web
Reachable by mini. Desktop Chrome and Firefox, iPhone Safari, Android Chrome.
Good questions.
Is this a real run?
Yes. mini explored our food-ordering demo app, wrote its 12 tests and ran them on iOS and Android earlier this year. The verdicts, the recordings and the cart bug are shown as that run produced them. It’s our own app, so no customer data is in it.
Why does the run show more than the cart bug?
mini reports every acceptance criterion that failed or looked off, as Critical or Warning, and it wrote each of those criteria itself, from what it learned exploring the app. This page follows the cart bug because it’s the easiest to check by hand. Open any other issue in the report and it carries its own recording and verdict.
What if mini flags something that isn’t a bug?
Mark it Not a bug with a short reason. mini uses the note on the next run, so the same false alarm isn’t raised again. If mini can’t grade a test, because the build broke or sign-in failed, it’s marked Unprocessable, never passed.
Can it run our hardest flows, not just a login?
Yes. One test can drive up to three iOS or Android devices, follow email codes, magic links and deep links out of the app and back, mock a location, or cut the network mid-run on Android and the web.Regression testing
Can we start from the test cases we already have?
Yes. Upload your existing test cases and specs, connect your repository, and mini explores the app. It writes its tests from all of it, and your team can edit any test in plain words.
Now bring your three hardest flows.
Bring a build. We set mini up and walk you through the run.