Production-readiness review

Is your Lovable or AI-built app ready for production?

It works in the demo. The question is whether it holds up once people pay for it.

You built your app with Lovable, Bolt, Cursor, Base44 or Replit. It works, people are signing up, and you are about to switch on payments, launch publicly or show it to investors. Before you do, you want to know whether it is ready, or whether it will fall over, leak data or charge someone twice in the first busy week.

The scans built into AI tools are worth running. Lovable, for example, checks that row-level security is switched on when you publish. But they answer a narrower question than the one in front of you. We read the app the way a careful engineer would before signing it off for real customers, and give you a straight answer: launch, fix these things first, or rebuild one part.

Planning a launch? Tell us the date and what you will charge for.

We review the app with that launch in mind and send you a ranked list of what has to change first, and what can wait.

Ask for a launch reviewWe reply to every enquiry within 24 hours.

What production ready actually means

An app is ready for production when it can serve real users and take real money without you watching it every hour. In practice that means five things are true: each customer's data is kept apart from everyone else's; payments cannot be faked, lost or doubled; you hear about failures before your customers tell you; data can be recovered after a mistake; and the next change, whether you, a developer or an AI tool makes it, cannot quietly undo the rest.

None of that shows up in a demo, which is exactly why apps that look finished are so often not ready. The checklist below is what we look at before we would call an app production ready.

The production-readiness checklist

  1. 1

    Each customer only sees their own data

    Row-level security on every table, permission checks on the server, and, if you sell to businesses, one company's records kept apart from another's in the database itself. This is the item that decides whether you can take business customers at all.

  2. 2

    No secret keys where visitors can read them

    Payment, AI, email and database admin keys live on the server, and any key that was ever exposed has been rotated.

  3. 3

    Payments that cannot be faked or doubled

    Access is granted from verified webhook events, not from the success page; the same event arriving twice does not credit twice; failed cards, refunds, upgrades and cancellations all leave the account in the right state.

  4. 4

    Sign-up and sign-in that hold up

    Email verification, password reset, sessions that expire, and account deletion that really deletes. These are the flows real users hit on day one and demos never exercise.

  5. 5

    A data model you can change without breaking live data

    Database changes kept as migrations in the repository rather than clicks in a dashboard, so the next feature does not mean editing production by hand.

  6. 6

    Performance with real amounts of data

    Queries with indexes, lists that page instead of loading whole tables, and images sized for phones. Apps that are instant with 50 rows are often unusable with 50,000.

  7. 7

    You find out when something breaks

    Errors are logged somewhere you look, and someone is alerted when sign-ups, payments or the site itself stop working, before customers email you.

  8. 8

    Backups you have restored at least once

    For the database and for uploaded files, with written steps for getting them back.

  9. 9

    Separate test and live environments

    A test database you can break, a live one you do not touch by hand, deploys that can be rolled back, and settings kept out of the code.

  10. 10

    The legal and operational basics

    A privacy policy and cookie consent that match what the app really does (essential if you have users in the EU), terms of service, and email set up so your sign-up and receipt emails reach the inbox instead of spam.

Want to check the security items yourself first? Our vibe coding security checklist shows how to test each one, and if Supabase is showing you a warning, start with what "RLS Disabled in Public" means. Built your app in Lovable? See the Lovable-specific checks.

Launch, fix first, or rebuild one part

Every review ends in one of three answers, and we tell you which one, plainly.

Go

Nothing blocks launch. You get a short list of things to improve over the next weeks, ranked, so you can launch now and tidy up as you grow.

Fix first

The common result. A handful of specific problems must be fixed before you take payments or personal data, usually configuration and missing checks rather than features. Days of work, and your users will not notice anything changed.

Rebuild one part

Occasionally one area cannot be made safe as it is: most often a data model that cannot keep customers apart, or a payment flow with no reliable record of who paid for what. You rebuild that area, not the app.

A full rewrite is almost never the answer. When someone recommends one before reading the code, they are guessing.

What usually blocks a launch

These are the problems we see most often in AI-built apps on their way to production. None of them is visible from the app itself; all of them matter as soon as real money and real data are involved.

  • Test and live data sharing one database, so a test run can change real customers' records.
  • Paid features unlocked by visiting the payment success page.
  • Webhook endpoints that accept requests without checking the signature.
  • Admin screens hidden from the menu but reachable by anyone who knows the address.
  • Tables with row-level security switched off, or with policies that allow everything.
  • An AI feature that anyone can call in a loop, with no limit and no spending cap.
  • No error tracking at all, so the first sign of a problem is an angry customer.
  • Uploaded files with no backup, because the database backup was assumed to cover them.

Apps we have taken to production ourselves

We can tell you what production needs because we build and ship production software for clients. Two examples:

  • Lead Platform, Belgium: a live renovation marketplace with three separate portals, Stripe credit purchases handled by an idempotent webhook with a guard against crediting twice, live EU VAT checks, and the whole interface in English, French and Dutch.
  • Run Courier, United Kingdom: a same-day courier platform where web, iOS and Android apps share one Supabase backend with row-level security, live GPS tracking, Stripe payments and six role-based dashboards that keep customer prices and driver pay apart.

How it works, and what it costs

  1. Audit, $149, back in 48 hours. A senior engineer reviews your code and infrastructure for access control, secrets, data and input, and production readiness, and you get a severity-ranked report with a fix for every finding. Tell us your launch plans and we point the review at them.
  2. Fix Sprint, typically $600 to $1,500, one to two weeks. We fix the critical and important findings, test them again and issue an updated report showing each one closed.
  3. Production hardening, from $2,500, scoped per project. Prototype to production: authentication, keeping customers' data apart, payments, monitoring, backups, and anything that genuinely needs rebuilding.

The $149 audit fee comes off the price of whatever you do next. Read-only access to your repository is enough to start, and we sign an NDA on request. The full audit, including what it is not, is described on the vibe coded app audit page.

Find out whether your app is ready before your customers do.

$149, 48 hours, and the fee comes off any fix work you choose afterwards.

Ask for a launch reviewWe reply to every enquiry within 24 hours.

Frequently asked questions

Is my Lovable app production ready?
It can be, but working is not the same as ready. Lovable, like other AI builders, gets you to a working app fast and now checks some security settings when you publish. What it cannot decide for you is whether your payments, customer data, monitoring and backups will hold up once real people depend on the app. The checklist on this page is how we answer that question for any AI-built app.
Can an app built with an AI tool handle real users and payments?
Yes. Many do. The tools generate perfectly usable products; the gaps are usually in the parts a demo never touches, such as access rules, webhooks, error tracking and backups. Those are fixable without starting again.
Do I need to rebuild my app before launching?
Rarely. In most apps the problems are configuration and missing checks, which are fixed in place. A partial rebuild is only worth it when one area, typically the data model or the payment flow, cannot be made safe as it is, and we will say plainly when that is not the case.
How long does it take to get production ready?
The audit comes back in 48 hours. Fixing the critical and important findings in a Fix Sprint usually takes one to two weeks. Larger production hardening is scoped per project, because it depends on what the review finds.
What does it cost?
The audit is $149 and that fee comes off whatever you do next. A Fix Sprint is typically $600 to $1,500 depending on the findings, and full production hardening starts at $2,500.
Can I keep building with my AI tool afterwards?
That is your choice. We work in your repository and tell you which parts to be careful with, such as access rules, payment handling and database migrations, so later changes, whoever or whatever makes them, do not undo the fixes.