BLOG
BlogGuides

Ten tests to run before you share your AI-built app

A pre-launch checklist for apps built with Claude Code, Cursor, Lovable or Bolt, written as ten plain-language tests you can run on the live address today.

A studio that ships apps built with Lovable, Bolt and Cursor wrote up the crash that changed its process: an app went down 18 hours after launch. Their estimate of the cost of auditing before launch: about 30 minutes per app.

Here are ten checks for that half hour, each written as a test Testsy can run against your live address. Replace the names and values with your own. Use a staging copy or made-up details where a test creates data.

The ten#

1. A stranger can sign up.

Sign up with this run's email address, confirm the account from the
email, and check the dashboard shows the new account.

2. The confirmation link goes to your real site. Add to test 1: "check the page the email link opens is on app.example.com". Why it might not.

3. Sign out works.

Sign in with the stored account "customer", sign out, press the
browser's back button, and check the dashboard is not shown.

4. What you save is saved.

Create a project named "Launch check", reload the page, and check
it is still listed.

5. Another account cannot see it. A two-person test: Test that one user cannot see another user's data.

6. Password reset works end to end. The full test.

7. The main pages render. One check per page that reads a heading or a value, with JavaScript errors set to fail. Catch the blank white screen.

8. It works on a phone screen. Run test 1 again with the screen set to mobile (412 × 914). Buttons hidden below a fold or menus that do not open show up here.

9. Empty states make sense. A brand-new account sees no data. Check the first screen says what to do next instead of showing "undefined" or an empty table.

Sign up with this run's email address and confirm it. Check the
dashboard tells the new user how to create their first project
and does not show the word "undefined".

10. Payments, if you take them. Go through checkout in your payment provider's test mode, with its published test card, and check the account shows the paid plan afterwards. Never on a live store. Testing Stripe checkout and plan upgrades.

Run them together#

Put the ten in one project and press Run all. The project page groups them by result, so you see what failed at a glance, and each failure carries what was expected, what was seen and a replay.

A project: tests grouped by their last result, with Run on each and Run all for the project.
A project: tests grouped by their last result, with Run on each and Run all for the project.

Ten tests is ten runs. The free trial has 10, so you can run the whole list once before paying anything.

After launch#

Keep the project. Run it after every change your agent makes (that is where regressions come from), and put the sign-up test on a nightly schedule: it depends on your email provider and auth settings, which change without you touching the code.

This post is part of Testing an app you built with AI.

Filed underWebAI-built appsReleases and regressions