How much does it cost to fix a vibe-coded MVP?
The short answer
Most rescues land between $8,000 and $30,000. The number is driven by how many features exist, not how complex any one of them is, because the work is replacing the layer underneath every feature: authorization, schema, deployment and backups. A three-to-five day assessment gives you a firm figure before you commit.
What drives the number
Not the thing people expect. Founders assume the price tracks how complicated their product is. It tracks how much surface area the foundational work has to touch.
- Feature count, because server-side authorization has to be added to every endpoint and pagination to every list.
- Number of user roles, because each one multiplies the permission rules that need writing and testing.
- Third-party integrations, because each is a place where secrets were exposed and failures were never handled.
- How much real data already exists, because a schema with constraints cannot be applied over rows that violate them until those rows are cleaned.
- Whether anyone else has already tried to fix it, which occasionally halves the work and occasionally doubles it.
Screen count barely matters. A five-screen product with four integrations and three roles costs more to rescue than a twenty-screen product with one role and no integrations.
The three bands
| Band | Typical range | What you are dealing with |
|---|---|---|
| Hardening | $8,000–14,000 | The product is sound. Authorization, secrets, schema constraints, deployment and backups, in three to four weeks. |
| Rescue | $14,000–30,000 | The above, plus meaningful refactoring: data model changes, replacing an integration that was never going to hold, breaking up code nobody can extend. |
| Rebuild the core | $30,000+ | The domain model is wrong in a way that cannot be migrated around. Rare. We only recommend it with specific evidence in hand. |
Roughly seven in ten of the codebases we assess fall in the first two bands. The instinct that a generated codebase must be thrown away is usually wrong, and it is an expensive instinct to follow.
Why an assessment comes first
Nobody can price this from a description, and you should be suspicious of anyone who tries. The variables above are only visible from inside the repository.
An assessment is three to five days of reading the code, running it, and deliberately attacking it, ending in a written report: what is there, what is broken, what it costs to fix, ranked. Ours is $2,500 and the report is yours whether or not you continue with us.
Comparing the alternatives honestly
| Option | Cost | The real risk |
|---|---|---|
| Do nothing | $0 now | The bill arrives as a breach, an outage during a launch, or a customer's data in the wrong hands |
| Cheapest available freelancer | $2,000–6,000 | Patches the symptoms you noticed; the foundation stays missing, and you pay twice |
| Rescue engagement | $8,000–30,000 | Real money, spent on work you cannot see, on a product that already appeared to function |
| Full rewrite | $40,000+ | Months with no new features, and a real chance the second version never ships |
The genuinely difficult one is the second row. A cheap fix feels proportionate because the visible problem is small. It usually is not the problem.