Questions, answered
Everything people ask before their first run. Not here? Write to us.
Product
Do I need to write any test code?
No. You describe what a user should be able to do, for example “sign up, add a project, invite a teammate”, and the agent works out the clicks in a real browser. There are no selectors to record and no scripts to maintain when your UI changes.
Which apps can it test?
Any web app reachable at a URL: production, staging, or a preview deploy. If a human could open it in Chrome and click through it, an agent can test it.
How is this different from Playwright or Cypress?
Those are frameworks: you write and maintain the tests. SanityCheck is the engineer. You state the goal in a sentence or two and it does the testing, including the parts scripts are bad at, like judging whether the page actually looks right.
Can tests run automatically when I deploy?
Yes. Connect GitHub, GitLab, or Vercel from the Integrations page and mark any test “run on deploy”. When a deployment succeeds on the branch or environment you chose, those tests launch against the new build on their own. CI/CD integrations are included on every paid plan.
What counts as a run?
One run is one AI agent in one browser executing one test once. Run a test across 5 browsers in parallel and that counts as 5 runs. Every run includes the live view and the full action, console, and network replay.
Security
Is it safe to give it my app's credentials?
Test credentials are encrypted at rest and filled into login forms by the system, never by the model. The AI never sees the password. You also control allowed domains and hard “never do” rules per application.
What happens when I hit my monthly cap?
Runs pause until your period resets or you upgrade. There is no metered overage. A bug that sends an agent looping can never turn into a surprise bill.
Support
A run failed and I don't think my app is at fault.
Open the run and click the failing item. The step timeline shows what the agent tried and why it reported a failure. If it still looks wrong, email the run link to [email protected] and we will review it.
The agent got stuck at a login or 2FA screen.
Add a login on the application page so the system can sign in for the agent. For a one-time code, the agent pauses and asks you during the run. Note in the login details if 2FA is required.
How do I change or cancel my plan?
Go to Settings, then Billing, and open the billing portal. Upgrades apply right away. If you cancel, the plan stays active until the end of the period you paid for.
Can you add a feature I need?
Tell us what you're trying to do. Deploy triggers for GitHub, GitLab, and Vercel came from a request like that.
Stop testing by hand.Not sure if it will work on your app?
Try it. 15 free runs a month, no credit card, and setup takes less than two minutes.