Security review

Security review for AI-built apps.

We review the security of apps built with Claude Code, Cursor, Lovable or Bolt: $350 for a written audit in 2 days, or from $750 with every fix applied and re-tested (prices from November 2026). We test by hand, logging in as two real users to try to read and change each other's data, then check for exposed keys, unprotected API routes and public file storage.

Audit $350, report in 2 days. Audit with fixes from $750.

  • Fixed price, agreed first
  • Keep building with your AI tool
  • Everything in your name

Sound familiar?

What AI tools tend to leave open

01

Data one user can read for another

Row level security off, or a policy of USING (true) that looks enabled while every row stays public. The most common hole we find, whichever tool wrote the code.

02

Keys shipped to the browser

Service keys, Stripe secret keys or OpenAI keys in frontend code or behind a NEXT_PUBLIC_ or VITE_ prefix, where anyone can copy them.

03

Admin checks in the UI only

The admin panel hides itself in React, but the API behind it answers anyone who calls it directly.

04

Public file storage

Invoices, ID documents and private uploads in buckets marked public, reachable by anyone with the URL.

Prices

Fixed prices, written down first

Security audit

Know where you stand

$350

  • Access rules tested with two real accounts
  • Exposed keys and secrets check
  • Auth, API routes and storage
  • Written report ranked by severity
  • Fee credited if you add the fixes

Audit + fixes

Found, fixed, re-tested

from $750

  • Everything in the audit
  • Every finding fixed in your repo
  • Secrets moved server-side and rotated
  • Re-test after the fixes
  • You keep building with your AI tool

The same published prices as our Lovable security audit, November 2026. Large apps (many roles, many tables or many API routes) get a fixed quote before we start. After the fixes, the AI Build Partner retainer (from $500 a month) keeps a developer reviewing what the AI changes next.

What you get

What you get

  • A written report ranked by severity, in plain English, with proof for each finding
  • Database access rules tested with two real accounts
  • API routes, server functions and edge functions checked for login checks, input validation and rate limits
  • With fixes: access rules rewritten, secrets moved server-side and rotated, admin checks on the server, storage made private
  • With fixes: a re-test after the changes, so you know they hold

Fixed price, agreed after a free call. No hourly surprises.

How it goes

  1. 01

    Free call

    Tell us what the app does, who uses it and which tool built it. We agree a fixed price.

  2. 02

    Access

    Invite us to the repo and the database project. We test with throwaway accounts.

  3. 03

    Audit and fix

    We test, write up and fix. Nothing changes in production without your OK.

  4. 04

    Report and re-test

    You get the report in 2 days, then the fixes and a clean re-test if you chose them.

“Lovable 2 Live has a real focus on quality. They're the team we trust on Klick.now when something has to work in production, and they're always straight with us about what's worth building and what isn't. Easy recommendation.”
Riccardo, Founder, Klick.now

Questions

Asked before every call

Who can do a security review of an app built with Claude Code or Cursor?

We do, as a fixed-price job: $350 for a written security audit in 2 days, with database access rules tested by logging in as two real users, plus exposed keys, auth, API routes and file storage, ranked by severity with proof. From $750 with every fix applied in your repo and re-tested. Checked November 16, 2026.

Is code written by AI insecure?

Not by default, but AI tools write what you ask for, and security rules are rarely part of the prompt. The holes we find most often are the same across Claude Code, Cursor, Lovable and Bolt apps: database rules that let any signed-in user read everything, keys in the browser, admin checks only in the UI and public storage.

Do you review Bolt and Lovable apps too?

Yes. Lovable and Bolt apps usually run on Supabase, so the review focuses on row level security, edge functions and storage buckets. For Lovable apps we also have a dedicated Lovable security audit page with a sample report.

How long does it take and what does it cost?

The written report arrives 2 days after we have access to the repo and the database. The audit is $350; the audit with every fix and a re-test starts at $750, and the $350 is credited if you start with the audit and then add the fixes.

Is this a penetration test?

It is a hands-on test of the attacks that matter most in AI-built apps: a second account trying to read and change the first account's data, users making themselves admin, API routes called without a login and private files reachable by URL. It is not a certified pentest report for compliance; if a customer asks for one, we say so on the free call.

Do you need my production data?

No. We test access rules with throwaway accounts and only look at real data where a finding needs proof.

Let's look
at your app.

Fifteen minutes is usually enough to tell you what's wrong and what it costs to fix.