Describe what to check in plain words. An AI tester runs it in a real browser or a phone simulator and shows you what it saw, step by step.
No password. The link signs you in, and creates your account if you're new. Free while in beta.
We sent a sign-in link to . Open it to sign in. It can take a minute to arrive, and may land in spam.
By signing in you agree to the terms and the privacy notice.
see how it worksThe heart turned red, but Favourites stayed empty.
No test framework, no recorder, no selectors. The tester reads the screen the way a person does, and tells you what it found the way a person would.
A title and a few numbered lines: what to do, what should be on screen. Attach a design image to a step when the pixels matter.
The helper app pairs your Mac with the site and checks the tooling for you. Websites run in Chromium, apps in the iOS Simulator or an Android emulator.
Every step gets a verdict and a screenshot. Failed means the app did something wrong; what the tester couldn't tell is marked inconclusive, with the reason. One click turns a failure into a bug report.
This is what a run looks like while it happens: the tester says what it sees, does one thing at a time, and flags the moment the app goes wrong.
A scripted test breaks when a button moves. A tester that reads the screen just finds the button, and notices what's wrong around it.
The heart turned red, but Favourites stayed empty.
The app refused the address and said so in plain words.
The map tiles never loaded on the emulator, so I couldn't judge the position marker.
One tester for all three. The helper app on your Mac does the setup; the site keeps the tests and the results.
Runs in a headless Chromium on your Mac. Give it a URL and, when needed, a test account.
A simulator build of your app in the iOS Simulator. Xcode on the Mac; the helper handles the rest.
An APK on an Android emulator. The helper checks the tooling and shows you what's booted.
An Anthropic, OpenAI or Google Gemini key, stored encrypted. Each test shows what it cost.
The things that make scripted UI tests expensive to keep are the things the tester does for you.
The tester reads the screen like a person. A moved button is still the button.
Per project, the tester writes down what it learned — where the cookie dialog hides, which build is on the device — and reads them before the next run.
Passed tests skip reruns until your app's version name changes. Clear a pass whenever you want it rerun.
Attach a design image to a step. If the live screen drifts from it, the test fails and shows you both.
Watch a run live. Type a hint — “the menu is behind the avatar” — and the tester takes it mid-test.
A failure becomes a bug report with the steps, the screenshot and the cause, ready to paste into Jira or anywhere else.
Every account starts on the free plan. Your model provider bills your key; the site itself costs nothing during the beta.
Sign in with your email and you're on it. No card, no trial clock.
A paid plan comes later. When the free period ends, your projects, tests and results stay yours to read and edit; only starting new runs will need a plan. The site will say so before it happens.
No. A test is a title and a few numbered lines in plain words. If you can brief a colleague on what to check, you can write one.
The site is free during the beta. The tester thinks with a model from a provider you choose — Anthropic, OpenAI or Google Gemini — using your own API key, and that provider bills you. Each test's result shows what it cost.
On a Mac of yours with the helper app installed (Apple silicon). Websites run in a headless Chromium there, iPhone apps in the iOS Simulator, Android apps in an emulator. The site queues the run and keeps the results.
A simulator build (.app) for iOS or an APK for Android, and the simulator or emulator installed on the Mac. The helper's Doctor screen checks what's missing and tells you how to fix it.
It's stored encrypted (AES-256-GCM) with a key that lives only in the API service, decrypted once per run, used only for your runs, and removable at any time under Providers & keys.
Your test text, its notes, and the text and structure of the screens it reads. Screenshots are kept for you and are not sent to the model. See the privacy notice for the full picture.
Your account keeps everything: projects, tests, notes and results stay readable and editable. Only starting new runs will need a paid plan, and the site will say so before it happens.
Sign in with your email, pair a Mac, add a key. The first run is minutes away.