170 AI-built apps with tables anyone could read
A researcher scanned apps built with Lovable and found 303 database endpoints across 170 projects without proper access rules.
Your AI wrote the code. It didn't lock the doors.
Apps built with Cursor, Lovable, Bolt or v0 keep going live with open databases and API keys left in the browser. We show you what an attacker sees, then hand you the fix as a prompt for your AI tool.
Three public cases since 2025. Same story each time: the app worked, and nobody checked what it exposed.
A researcher scanned apps built with Lovable and found 303 database endpoints across 170 projects without proper access rules.
The dating-safety app Tea left a Firebase storage bucket readable by anyone. About 13,000 of the images were selfies and photo IDs.
The Supabase key sat in the site's JavaScript and the database had no row-level security. The founder said AI wrote all of the code.
These companies are not NodaGuard customers. They are publicly reported cases of the exact mistakes our scan looks for.
A working app and a safe app are two different things. Your AI tool only promises the first one.
/.env or /.git?Nobody stands between those two lists. That gap is what NodaGuard checks.
No install, no security knowledge needed. If you can paste into your AI tool, you can fix what we find.
Pick a plan, then give us your address. Nothing to install and no code added to your app: we look at your live site from the outside, the way an attacker would.
https://yourapp.comA 0–100 security score and every finding, sorted by how dangerous it is.
score 38 · 2 critical · 1 highEach fix comes as a ready-made prompt for Cursor, Claude or Lovable. Every finding has one.
re-scan → score 96Deeper checks (open databases, exposed files, unprotected APIs) run only after you prove the site is yours. An automated scan does not replace a penetration test.
Scan My AppThe most common problems for each setup. Click one to see what can go wrong and how to fix it.
Most common problems (sample)
Your database answers anyone who asks. No login needed, so one request pulls every user's email and phone number.
Turn on row-level security so each user only sees their own rows. You get the exact SQL and a prompt for your AI tool.
Automated scanners probe new sites around the clock. Here is what that looks like, and what changes after the fix.
Automated scanners test every new site for the same few mistakes. A sample of the first five minutes:
One scan, a few pasted prompts, a second scan.
Hover a card, or press the button, to see what a scan finds behind it.
Yes. Here is what happens, in order.
Pay once, get every fix. No hidden subscription trap.
A full scan of one site: your security score and every issue we find, sorted by severity. For each issue you get what it is, why it's dangerous, how to fix it, and a ready-made prompt to paste into your AI tool, plus a PDF report. After you fix things, one re-check scan within 30 days is free.
Payments are processed by Paddle. Your card details never reach us and we don't store them. Credits are a one-time payment and never auto-renew. Your invoice is emailed to you automatically.
No. For a web scan you just enter your address. For source code you pick your project folder (or a ZIP of your repo), and for an Android app you pick the APK file; both are checked right in your browser. No library or code is added to your project.
No. The scan doesn't delete or change data, doesn't submit forms, and keeps requests per second low. Your users won't notice a thing.
No. Before a scan you prove the address is yours with a DNS record or a small file you place on the site. Unauthorized scanning is illegal and against our rules.
No. Your APK and your source code are opened and checked inside your own browser, on your device; the files are never sent to us. Only the list of findings is saved to your account so you can reopen the report, and if a key is found we keep a masked sample, never the key itself. For source code we also look up the names and versions of your dependencies in a public vulnerability database.
For web scans the stack doesn't matter: Next.js, React, Vue, Laravel, Django, WordPress or anything else. Supabase and Firebase get additional dedicated checks. Source-code scanning supports JavaScript, TypeScript, Python, PHP, Go and Java.
Not entirely. An automated scan catches the mistakes attackers try first and exploit most often. Complex flaws specific to your business logic need an expert eye. For most small teams, an automated scan is the right place to start.
Scans you buy are valid for one year from the day you pay. They never renew on their own.