Apps built with Lovable, Bolt, v0, Cursor or Claude
Your app works.
But is it safe?
We check exposed keys, database rules, data across accounts and payments. We explain what we find in plain language, and we fix it.
→ scanning public JavaScript…✕ Supabase service_role key visible in the browser→ probing the database with the public key…✕ «profiles» readable without login · 1,240 rows→ testing with two test accounts…✕ account A sees 2 orders from account B⚠ users can change their own is_admin✓ private notes: protected✓ findings verified by a person · report ready
Send us your link. Within 24 hours, privately. We only look at what any visitor can see.
Why now
AI tools build fast. They also repeat the same mistakes.
growth in new monthly vulnerabilities linked to AI-generated code, January to March 2026.
vulnerabilities and 400+ exposed secrets across 5,600 public AI-built apps.
of public apps on one popular platform were leaking user data.
Sources: Cloud Security Alliance (April 2026), Escape.tech, and public reports covered by the security press.
What we check
The four issues we see most in AI-built apps.
Keys in the page
Supabase's master key, or Stripe, OpenAI or Anthropic keys anyone can read.
Database rules
Tables without RLS, open Firebase databases, public files.
One account, another's data
With two test accounts: can one see or change the other's data? Can a user make themselves admin?
Payments and AI
Unsigned webhooks, prices taken from the browser, unlimited AI usage that can cost you thousands.
How it works
Outside check, free
You send your link and we review what any visitor receives. We reply privately.
Full review, with your written permission
You sign an authorization, give us two test accounts, and we review the database, permissions, payments and AI.
Fix
We fix what's critical and re-test every item.
Ongoing
Weekly scan and a review after every major change.
What a report looks like
Agenda Bella, a practice app with made-up data.
Names, emails and phone numbers for anyone. Fixed by enabling RLS with per-user policies: 1 hour.
Confirmed with two test accounts. The policy said using (true): 30 minutes.
By changing a field on her own profile. Tested and reverted.
RLS on with per-user policies. Re-tested without login: 0 rows.
Owner-only policy. Re-tested with both accounts.
Only the server changes roles. Repeated attempt: rejected.
Each customer can only read and write her own.
Pricing
Outside check
Exposed keys, public files, headers.
What's included
- We only look at what any visitor's browser receives: no logging in, no touching your database.
- Secret keys visible in the JavaScript: Supabase's master key (service_role), Stripe, OpenAI, Anthropic, AWS or Resend.
- Published source maps that expose your original code.
- Your site's security headers and which services your app uses (Supabase, Firebase).
- Delivery: a private message within 24 hours with what we found and what to fix first.
- We need: just your app's link.
Full review
Small apps. Report plus a 30-minute call.
What's included
- Everything in the outside check, plus:
- Database rules: every Supabase table with RLS and per-user policies, and your Firebase rules.
- Two test accounts: whether one account can see, change or delete the other's data.
- Whether a user can upgrade their own role or plan, or give themselves credits (tested and reverted).
- Forgotten files on the server (.env, .git) and server functions that don't check who's calling.
- Payments: that your Stripe webhook verifies the signature and prices are computed on the server.
- AI inside your app: keys only on the server and a per-user usage limit.
- Delivery: a plain-language report (what could happen, how we confirmed it, how to fix it) and a 30-minute call. Critical issues are reported the same day.
- Timeline: 3 to 5 business days from the signed authorization.
- We need: the signed authorization, two test accounts and, optionally, read access to your Supabase or Firebase project.
- Scope: apps up to about 15 tables; if yours is bigger, we quote first.
Fix
A fixed price before we start.
What's included
- We fix the critical and high items from the report: database policies, keys moved to the server and rotated, protected roles and credits, webhooks and AI limits.
- We re-test every item and send you the updated report.
- A fixed price in writing before we start: no hourly billing, no surprises.
- We need: access to your project (Supabase, Firebase or the repository) during the fix.
Ongoing
Weekly scan and a review after every major change. Cancel anytime.
What's included
- Weekly outside scan to catch exposed keys or files published by mistake.
- A security review after every major change to your app (up to 2 per month).
- Immediate alert if anything critical shows up.
- Monthly summary of what was reviewed.
- Billed monthly through PayPal; cancel anytime from your own account.
- Best after the full review; if you haven't done it, we start there.
Questions
I'm not technical. Is this for me?
Yes: the report is written for whoever built the app, not for an engineer. Each item says what could happen and how to fix it.
Doesn't my builder handle security?
Partly. The platform secures its infrastructure, but your database rules, who sees what, and how you charge are your app's decisions, and that's where most issues are.
What do you need for the full review?
The signed authorization, two test accounts and, optionally, read access to your Supabase or Firebase project.
Is it reviewed by AI or by a person?
Both. We use tools and AI to review faster, and a person reproduces every finding, discards false positives and signs the report.
AI-assisted, human-verified: every finding is reproduced and signed off by a person who stands behind it. AI is a support tool and does not replace professional validation.