By Keep prompting while each attempt moves you forward and the error lives in one screen. Get a human when the same bug has survived three fixes, when a fix breaks something unrelated, or when the problem touches data access, payments or sign-in.
Every builder says the same thing when something breaks. Try to fix. Lovable's own troubleshooting order is Try to fix first, then Chat or Plan mode, then revert the last change. That order is right for a visible error in one place. It is wrong for the fourth attempt at the same bug, and the builder will not say when you crossed that line.
Signs that prompting has stopped working:
- The AI says it is fixed and it is not.
- Each fix breaks something that worked yesterday.
- The error has moved. It started in checkout and is now in sign-in.
- The problem is on the live site only. The AI can see the preview. It cannot see your host, your custom domain or your production database.
- The question is who can see what. Access rules, keys and webhooks fail silently. The screen looks right while the data leaks.
- You have reverted to a working version more than once this week.
Lovable gives you 10 free Try to fix runs, each coming back 24 hours after use. After that every attempt uses credits, and credits are non-refundable, including the ones spent on fixes that did not work. Plan mode costs one credit per message. Lovable values a credit at $0.25 on Pro and $0.50 on Business. A single human fix costs €99–$990 and takes about two days. If three days of prompting have cost you more than the low end of that, the human is already the cheaper option. A Replit forum member with months of builds behind them put it plainly: the last 10% is significantly more expensive than early building.
A simple rule we use. Three strikes on one bug, or one fix that breaks something unrelated, and you stop prompting and get a person. Anything touching sign-in, payments or another user's data, you get a person on the first failure.
What to do before you hand it over, so the human's first day is not spent on setup:
- Connect the project to GitHub. The developer needs a copy they can read without using your credits.
- Revert to the last version that worked, and note which one it is.
- Write down what you asked for, what the AI said it did, and what actually happened.
- Note which backend you use, Lovable Cloud or your own Supabase, and whether payments are live.
Money and time. A check or audit is $49–$800 and takes a day to a week. A single fix is €99–$990, about two days. If the list is long, a cleanup is $1,500–$15,000 over one to eight weeks. See pricing and the checklist.
Get help on the first failure, not the third, when the app already has paying users, when a key may have leaked, or when the builder itself suggests an outside developer. One Replit user was told exactly that by the Agent in October 2026, and a forum reply found the real cause in the data. Who can fix it lists your options.
Check these
The code lives in a Git repository you own
Connect GitHub or export the project. If the builder account is lost, the product should not be.
Access rules are on for every table
In Supabase, open Table Editor: every table should show RLS enabled, with at least one policy. In Firebase, your rules should never say allow read, write: if true.
Paid accounts are unlocked by a verified payment webhook
Access should change only when your server receives a signed webhook from Stripe or your provider, not when the browser reaches a success page.
Errors reach you before users do
Add error monitoring and uptime alerts. If you learn about outages from customers, this is missing.
Top-rated companies
Ranked by our score: published prices, scope of work, AI tools named, timing and a low-risk first step. How we score.
- 1Afterbuild Labs10.0/10Best for: A check before launchFree first step · from $49
- 2Redlab10.0/10Best for: A full rescue by one team that keeps buildingFree first step · from $490
- 3Vibe Code Rescue (Lev Arakelyan)9.0/10Best for: A one-off check and a cleanupFrom $299
Related questions
Will a developer just use AI to fix it anyway?
Often yes, and that is fine. The difference is that they read the result, test it against the rest of the app, and know what the AI cannot see, such as your host settings and production database.
Can I go back to prompting after a human has fixed it?
Yes. Ask for the fix to be made in the GitHub repository so it syncs back into Lovable. Ask for a short note on what changed and why, so your next prompt does not undo it.
Sources: Lovable docs: FAQ (Try to fix, refunds); Lovable docs: Plans and credits; Replit forum: Struggling to get past the last 10% of app builds; Replit forum: Replit told me to hire an outside developer.