Testing for apps built with AI
We sign in as several test users, walk through your app and show you the evidence: what breaks, what leaks between users and what doesn't add up. With a prompt to fix it in your AI tool.
Read-only access to your code · We never touch your real data
My clients
My clients
Pedro shouldn't see Ana's clients.
01The problem
Source: Symbiotic Security, analysis of 1,072 apps built with AI tools (2026).
02What we check
Every important flow completes, and what you saved is still there after a reload.
User B can't see or change User A's data.
A salesperson can't do what only an admin can, not even by going around the screen.
What the screen shows matches what was saved, and totals and statuses agree across screens.
Exposed private keys, tables open to anyone and dependencies with known flaws.
03How it works
With read-only access to your repository. You don't write tests or create users.
We rebuild your app in an isolated environment with fictional data and use it the way your users would.
What worked, what failed and how to reproduce it, with screenshots and a prompt to fix it with your AI. Then press “Test again”.
04Sample report
Fictional data. Every finding includes steps, expected versus observed, the evidence and the prompt to fix it.
See the full reportAny agent can see another agent's entire portfolio and change their sales stages.
Expected
Pedro only sees and edits his own clients; the API rejects the change.
Observed
Pedro's list includes Ana's clients and the API accepts the change (204).
PATCH /rest/v1/clients?id=eq.c_01… · Pedro
204 No Content · stage: "negotiation" → "closed"05Who it's for
“Does what I built actually work?”
You built with Lovable, Bolt, v0, Replit, Cursor or Claude. Before opening it to customers, check that saved data stays saved and nobody sees anyone else's.
“Did I get something that works?”
You paid a freelancer or an agency. Check the app you were handed before you launch it or make the final payment, without reading code.
06Pilot pricing
Limited spots during the pilot.
Free
Compatibility, exposed keys and open tables, in seconds. Without running your app.
US$9
Full review with test users and an evidence report, plus one re-run after you fix things.
Request my reviewComing soon
Re-tests automatically every time you change your app.
If we can't run the review because of something on our side, your credit is returned automatically. If configuration is missing, we tell you exactly what and nothing is used up.
07FAQ
No. We rebuild your app in an isolated test environment with fictional data. We never write to your real database.
Read-only access to your repository. We can't change your code; you make the fixes with the prompt we give you.
No. No review can guarantee that. We show you what we tested, what we found and what we couldn't check.
Web apps built with Lovable, Bolt, v0, Replit, Cursor or Claude that use React or Next.js with Supabase or Prisma. If your app isn't on GitHub yet, we'll guide you.
The quick check, seconds. The full review takes longer: we show progress and email you when the report is ready.
08Request your review
Tell us about your app. We'll reply by email with next steps; there's nothing to pay yet.