Production Ready
Make your Lovable, Bolt or Cursor app production ready
For apps you built with AI. The last stretch, done by people who ship.
Last reviewed
You built it with an app builder like Lovable, Bolt, Replit, Base44 or v0, a coding agent like Claude Code, Codex, Cursor or GitHub Copilot, or code from ChatGPT, Claude, Gemini or Grok. It works, until real users, real payments or a real attacker show up. Production Ready is a review and fix service. Within a week, a senior team finds what could leak data, lose money or break. Then we fix it in your code, and can run it after. You keep the app, the code and the momentum.
Sound familiar?
- "It works, but it's held together with duct tape and super glue." Every new feature breaks two old ones, and the AI keeps looping on the same bug.
- "I got a warning from Supabase." "RLS disabled in public," an email from a security researcher, or a news story about the tool you built with.
- "Our bill doesn't make sense." AI or cloud costs jumped, and nobody can say why.
- "An investor wants a technical review." Or an enterprise customer just sent a security questionnaire.
- "I can't read my own codebase." And you're about to hire your first engineer.
What goes wrong, and why
AI tools are great at getting you most of the way, fast. The last stretch, security, edge cases and upkeep, is where apps break. Google engineer Addy Osmani calls it the 70% problem. The same five problems show up again and again:
- Data open to anyone Supabase tables with row-level security off or written wrong. In 2025 a researcher found 170 of 1,645 apps on Lovable's showcase letting anyone read user data (CVE-2025-48757).
- Secret keys in the browser In January 2026, Moltbook left its database key in browser code with row-level security off, exposing 1.5 million API tokens (Wiz).
- Logins that don't really check In July 2025, a flaw let anyone register on private Base44 apps and get past login. It was fixed within a day (Wiz).
- Agents with too much power, and no backup In July 2025, Replit's agent deleted SaaStr's live database during a code freeze (Fortune).
- Runaway bills In August 2026, METR disclosed that a stolen key ran up about $600,000 in AI credits over three weeks. There was no spending cap on it (METR).
See the full dated list of incidents, or run our plain security checklist on your app today.
How we work
- Review A senior read of your app: a half day to two days of senior time, delivered within a week. We check data access, keys, logins, payments, backups, costs, hosting and code health. You get a ranked list in plain words, a fix plan, and an honest call on whether to rebuild or refactor.
- Sprint We fix what's dangerous first, then what's fragile: data locked down, keys moved to the server, real login checks, payments that charge, backups and monitoring. Work lands in your repo from day one.
- Takeover We move the code, keys and accounts into your name and run it: releases, backups, monitoring and cost limits, one or more days a week. We can help you hire your first engineer, and hand over a clean, documented codebase.
- Advisory You keep building with AI, safely. We set up the guardrails, review steps and spending limits so new features don't break old ones. It pairs with Claude for Your Team.
What the Review costs: we quote it before we start, based on the size of your app, and every dollar counts toward the Sprint if you go on to fixes.
Why us
We're pro AI. We built our own product with it. In 2026 we built and shipped Deliberate with Claude Code: a signed Mac app with automatic updates, a Rails registration and billing site, and integrations with Gmail, Outlook, Slack, Asana, Zoom and Granola. That is exactly the last stretch: logins with Google and Microsoft, billing and licensing, update delivery, a release process and privacy rules.
Before that: Payment Scholar, a venture-funded registration and payments platform for K-12 school districts, with student and payment data to protect; AthleteProgress, a video coaching platform that served ski clubs from around the world; and an operating system for solar installers. When the whole team is involved, we bring a combined nearly 55 years in startups and rapid-growth tech.
We treat every project as if it's headed to enterprise production and millions of users, even when it isn't. And we take two clients at a time, so you get senior attention.
Learn more
- Is your AI-built app secure? A plain ten-point checklist you can run today.
- Lovable, Bolt or Replit to production Stay on the platform or move off it, step by step.
- Rebuild or refactor? How to tell, and a five-minute self-test.
- AI-built app failures: a dated list The public incidents behind all of this, with sources.
- "RLS disabled in public" What Supabase's warning means, and what to fix today.
- Works in preview, breaks when published The usual causes, by symptom, and how to fix each.
Common Questions
Frequently asked questions
Is my Lovable app secure?
Lovable the platform and your app are not the same thing. Your app is only as safe as its data rules, keys and login checks. Check that row-level security is on, and correct, for every Supabase table; that no secret keys sit in browser code; and that logins and paywalls are checked on the server. Lovable has had platform problems too: CVE-2025-48757 in 2025, and in April 2026 a flaw that left some projects' code and database keys readable for 48 days.
Can I keep using Lovable, Cursor or Claude Code after you fix it?
Yes. We use AI tools every day. We set you up to use them safely: your own repo, a test copy of your data, review before changes go live, and spending limits on every key.
Should I rebuild or refactor?
Usually refactor. Your current app is the best spec of what you want. We decide after the review, based on whether the data model is sound, whether logins can be fixed, and whether tests can be added. You get an honest call either way.
Can I move my app off Lovable, Bolt, Replit or Base44?
Yes. The code moves through GitHub, the data is exported, secret keys are re-created by hand, and hosting and your domain move to accounts you own. Lovable's own docs note there is no one-click migration from Lovable Cloud, so it pays to plan it.
How long does it take?
The review takes a half day to two days of senior time, delivered within a week. Fixes take a few weeks, and the dangerous ones ship in the first week. A rebuild, if it comes to that, usually takes one to three months.
How much does it cost?
We quote the Review before we start, based on the size of your app: a half day to two days of senior time. Every dollar counts toward the fixes. After it, you get a written plan and a quote for the fixes before any work starts. For comparison, other sellers' own pages, checked November 17, 2026, listed reviews from $300 to $3,000, fix sprints from $1,500 to $20,000, and rebuilds from $15,000 to $150,000.
Who owns the code?
You do. Everything lands in your repo and your accounts. No lock-in.
My investor wants a technical review. Can you help?
Yes. We write a plain report that answers the questions investors and buyers now ask: how much of the code AI wrote, whether it is tested, how data is protected, the risks, and the plan to fix them.
What do you need from me to start?
The app's link, the tool you built it with, read access to the code or project, what is broken, and what worries you most.
Proof
How we built Deliberate with Claude Code
Deliberate is our own Mac app, built by a three-person team with Claude Code. We built the app, its registration and billing site, and integrations with Gmail, Outlook, Slack, Asana, Zoom and Granola. It runs on your Mac, never saves the full text of your email, and sends nothing without your yes.
Meet DeliberateWant a senior read on your app?
Send us the link and what worries you. We'll tell you what we'd fix first.