Data access
What we check
- Row-level security
- Storage policies
- Tenant isolation
- Queries scoped by user
- Admin functions


Ruben Roubish will see you on Monday 12 October at 11:30. The invite with a Google Meet link is on its way to you@company.com.
Ruben Roubish will confirm Monday 12 October at 11:30 by email at you@company.com shortly.

We reply within one business day. Prefer email? info@fourmeta.com
Thanks for reaching out. Ruben Roubish will get back to you at you@company.com within one business day.

01App audit[ Works with ]
Apps built with any AI builder or coding agent
We work with apps built on every major vibe-coding and AI platform
Since 2017, Fourmeta has designed and engineered digital products from London.
Now we pair that experience with AI workflows and a dedicated AI-built product audit to help founders turn Lovable, Bolt, Replit and Cursor apps into products that are secure, scalable and ready for real users.
AI builders nail the happy path. Real users, payments and due diligence find the rest.
signs your app needs an audit
Recognise two or more?
Database rules never tested as another user
Keys in the browser or in Git history
Login works, but the edge cases do not
Every new prompt breaks something else
No tests, alerts or restored backups
An investor asked a question you could not answer
Lovable projects left open
critical vulnerabilities
AI code tasks with a flaw
deleted by an AI agent
Research numbers come from specific samples and methods, not a defect rate for your app. That is what the audit is for.
What we check
Code, database, config and the live app, with a second test account trying to get in.
What we check
What we check
What we check
What we check
What we check
What we check
What we check
AI builders nail the happy path. Real users, payments and due diligence find the rest.
A scanner gives you 300 warningsYou need to know which five stop your launch
Works and should stay as it is
Must be closed before launch or payments
Works today, slows every new feature
Repair is slower or riskier than replacement
Not needed at this stage
Final recommendation, one of four
Written on the first page of the founder summary, next to the top five risks and the effort to fix
// How it works
You own your code at every step. Keep building in your AI tool, or move to a standard stack if the audit says so.
A 15-minute call or a short form. Scope, stack and fixed price. NDA on request.
Read-only access to the repo, the backend and a staging or live URL.
Two test accounts, real flows, code and config review, data-access tests.
The written report and a call to agree the safest, fastest next step.
Fixed prices, agreed before we start. No hourly meter.
£490fixed price
Get my app auditedWhat you get
Next step
Time
£4,900fixed price
Get a fix quoteWhat we do
What you get
Time
from£8,000
Talk about a rebuildWhat we do
What you keep
Time
£1,500/ month
Get in touchWhat we do
When
The audit fee is credited in full if you go ahead with the fix within 30 days
How much it costs to fix an AI-built appRun the built-in scans, then let a senior engineer test what they cannot
[ Our work ]
We audit AI-built products because we build them.
Need product logic, UX and team readiness reviewed too? See our full product and UX audit
Good fit
Not a fit
[ FAQ ]
Yes. We work directly with Lovable projects and the Supabase backend behind them through GitHub. We close the launch blockers first — data access, exposed keys, auth — then the issues that slow you down. Depending on the audit, you can keep building in Lovable afterwards or move the code to a standard React and Node stack that any developer can maintain.
It can be, once it has been reviewed like any other production app. Public research shows the same gaps again and again: database rules that let one user read another user’s data, keys shipped to the browser, and no tests or alerts. If you store personal data or take payments, check these before launch, not after the first incident.
Two working days for most apps, counted from the moment we get read-only access. Larger products with several user roles, many integrations or more than one codebase can take longer; we confirm the timeline and the fixed price before we start, so there are no surprises midway.
The audit is a fixed £490. If you want us to fix what we find, our Vibe-to-Production sprint costs £4,900 and closes the launch blockers in two to three weeks. A full rebuild is rarely needed; when it is, it is scoped as an MVP project.
Read-only access to the code repository, the database or backend project (for example Supabase or Firebase), and a staging or live URL with two test accounts. We never ask for passwords in chat and we do not write to production during the audit. We are happy to sign an NDA before you share anything.
Usually not. Most findings in AI-built apps are fixable in place: tighten database policies, move keys to the server, add checks on the server side, add tests on critical paths. We recommend rebuilding only the parts where repair would be slower or riskier than replacement, and we explain why for each one.
No. The audit covers the security, code and product risks that most often block a launch, tested with real accounts against your real app. A formal penetration test or compliance certification is a separate engagement; if your customers or investors require one, we will tell you and help scope it with a specialist.
Any. We review apps built with Lovable, Bolt, Replit, Cursor, v0, Claude Code, Windsurf, Firebase Studio and GitHub Copilot, as well as mixed human and AI codebases. We review the product and the code you ship, not the logo of the tool that generated it.
Tell us about your app. We reply within one business day with scope, the fixed price and a start date.

Ruben Roubish
Chief Sales Officer