サービス一覧

We fix and extend apps built with AI, so they are secure and ready for real users

Keep what works, fix what is risky, before real customers rely on it.

Tools like Lovable, Bolt, Cursor and Claude Code get a first app built very quickly. Before it holds customer data or takes payments, there is a set of work you cannot see on the screen. DNA Maker reviews and fixes apps built with AI, with our own engineers in Bangkok and no subcontractors. We read the code, the database and the settings, tell you plainly what can stay and what has to change, then make the system secure, recoverable and maintainable.

Fixing apps built with AI: An engineer points at code on a laptop for a founder at a desk, under a Production check panel with Security, Backups and Load test ticked

We fix apps built with Lovable, Cursor, Claude Code and Bolt

Each kind of tool leaves different problems behind, so we start where that tool usually slips. Built with something else? Send it too.

  • Lovable

    Builds web apps from a chat and often uses Supabase as the database. We check each table’s access rules (Row Level Security) and any keys that ended up in browser code.

  • Bolt

    Builds the front end and back end in the browser. We move the code into your own repo, take settings out of the code, and set up a server and database someone can maintain.

  • Cursor

    A code editor with AI help. The code is already yours, but it often grows faster than its structure. We tidy the structure, add tests and set up code review so your team can keep using AI safely.

  • Claude Code

    An AI that writes and edits code from the terminal, across many files at once. We review what it changed: secrets, sign-in, user permissions and any packages it added.

What usually breaks when an AI-built app meets real users

When you test alone, everything looks fine. These problems show up with real users, real data and people looking for holes.

  • Security holes

    APIs that never check who is calling, so one user can view or change another user’s data.

  • Secrets in the code

    Keys for payments, email or AI sit in browser code or were committed to the repo. Anyone who sees them can use them.

  • Sign-in and permissions

    Password reset does not work, sessions never expire, or admins and customers are not kept apart.

  • No backups

    No backups, or backups nobody has ever restored. The day the database breaks, you do not know how much comes back.

  • Costs that run away

    Features that call AI or fetch data on every page load, with no cap, and nobody sees the number until the bill arrives.

  • Slower as users grow

    Queries without indexes, or the same data fetched again and again on one page. Fine with tens of rows, very slow with tens of thousands.

  • No tests

    Ask the AI to fix one thing and another breaks, because nothing checks that booking or payment still work.

  • No monitoring

    The system goes down or throws errors and nobody knows until a customer says so.

Keep what works. No need to rewrite everything.

Our first job is to sort out what is already good. Many apps have screens and flows users like, real data that has to be kept, and business logic you thought through. We keep those, and fix or rewrite only the risky parts.

Usually kept

  • Screens and flows your users already know
  • Existing customer and business data, moved over without loss
  • Business logic that already works correctly

Usually fixed or rewritten

  • Anything that touches sign-in, permissions and personal data
  • Payments and connections to outside services
  • A database structure that will not cope with more data

Send us your repo

Six steps from the day you send the code to the day the system is ready for real customers. You see the review, the scope and the price before you decide to hire us.

  1. 01

    Tell us what the app does

    Which tool built it, which database and services it uses, whether it has real users yet, and where you want it to get to.

  2. 02

    Give us read access

    Invite us to the repo on GitHub or GitLab as read-only, under an NDA. At this stage we do not need passwords to the live system.

  3. 03

    An engineer reviews and writes it up

    We read the code, database and settings, then list the findings by risk: what to keep, what to fix and what to rewrite.

  4. 04

    Talk it through and set the scope

    You talk to the engineer who did the review. Then we put the scope, a fixed price and delivery dates in a quotation before work starts.

  5. 05

    Fix it on a copy of the system

    Security and customer data come first. The fixes happen on a staging copy, so the current app stays open in the meantime.

  6. 06

    Test, go live and hand over

    We test before going live, then hand over the code and documents. Keep us on, or have your own team take it from there.

Example

What before and after look like

A fitness-class booking app a founder built with Lovable and Supabase, with real members already using it.

Before

  • Payment and AI keys sit in browser code, visible to anyone who views the source.
  • The bookings table has Row Level Security off, so any signed-in member can read everyone’s bookings.
  • Automatic backups exist, but nobody has ever tried a restore.
  • Class suggestions call the AI on every page load, with no spending cap.
  • No tests, and outages are found when members message to say so.

After

  • Keys moved to the server side, and the leaked set replaced with new keys.
  • Row Level Security on every table, so members see only their own bookings.
  • A real restore tested, with the steps written down for the team.
  • Suggestions cached and reused, a cap on AI calls and an alert when costs look unusual.
  • Tests for booking and payment, and the team alerted the moment an error happens.

The screens, the booking flow and all member data stayed as they were, so members carried on as before.

A made-up example to explain the work, not a real client.

A team that writes code with AI too

We do not see AI-built apps as the wrong way to build. We use AI to develop faster ourselves, with engineers making the calls on architecture, security and code quality. Every change goes through code review and testing before it goes live. Still at the idea stage? See MVP development for startups. Need a whole new system? See custom software development.

FAQ: fixing apps built with AI

Do you have to rewrite everything?

Usually not. We keep the screens, data and logic that already work, and fix or rewrite only the risky parts, such as sign-in, data access and payments. If a part is not worth fixing, we say so when we send the review.

Which tools do you work with?

Lovable, Bolt, Cursor, Claude Code and any other tool that produces code as files. The code just needs to be in a repo or exportable. If it is still stuck inside the tool, we help move it into your own repo first.

If we send you our code, will it and our idea stay confidential?

We work under an NDA on every project. For the review we ask for read-only access to the code and do not need passwords to the live system. The fixed system is designed along PDPA lines.

Does the app have to go offline while you fix it?

No. The fixes are made on a staging copy and moved to the live system once they pass testing. If we find a hole where customer data is leaking right now, we tell you at once and close that first.

How long does it take and what does it cost?

It depends on the size of the app and what the review finds. After the review we put the scope, a fixed price and delivery dates in a quotation before work starts. For a large job we offer a smaller first phase, starting with security.

Can we keep building with AI afterwards?

Yes. We tidy the code structure, add tests and write documentation so you or your team can keep adding features with AI, with the tests catching the times the AI breaks something that used to work.

Who looks after the system afterwards?

The team that fixed it can look after it, both fixing issues and building the next features. If you have your own team, we hand over the source code and documents so they can take over, and you are not tied to us.

Already have an app built with AI? Let our engineers look first.

Tell us what your app does and which tool built it. Our engineers will reply with how the review works and what to check first.

Tell us about your app