Lovable, Bolt or Claude Code prototype to production in 2026: $8,000 to $75,000
TL;DR Building an app yourself with Lovable, Bolt, Replit, Base44 or Claude Code costs about $20 to $100 a month in subscriptions. Taking that prototype to production costs $8,000 to $75,000 at US agency rates in 2026, from fixing the app you have to rebuilding its core while keeping the screens and flows. A review of the code first, usually $2,000 to $6,000, honestly tells you which of the two you need.
- Lovable, Bolt, Replit, Base44 and Claude Code plans start at about $16 to $25 a month, and the real cost of a prototype is your time.
- A production review of the code takes about 20 to 40 hours, so roughly $2,000 to $6,000 at US agency rates.
- Fixing and launching the prototype you have takes about 80 to 200 hours, so roughly $8,000 to $30,000.
- Rebuilding the core while keeping the screens and flows takes about 200 to 500 hours, so roughly $20,000 to $75,000.
- The usual gaps are database rules, payments, secret keys, a separate test copy with backups, and anything watching the app once it is live.
In this article
- What it costs to build an app with Lovable, Bolt, Replit or Claude Code
- What an AI-built prototype is usually missing
- Keep it, fix it or rebuild it?
- What taking it to production costs
- Where the hours go when you fix a prototype
- How long it takes
- Should you just keep prompting?
- Who owns the code?
- Can you keep using the AI tool afterwards?
- What to prepare before you ask for a quote
- When we are not the right fit
- How to get a number for yours
Building an app yourself with Lovable, Bolt, Replit, Base44 or Claude Code costs about $20 to $100 a month in subscriptions while you build, and taking that prototype to a production app real customers can rely on costs $8,000 to $75,000 at US agency rates in 2026.
The low end is fixing the prototype you already have, so sign-in, database rules, payments and a proper launch, and the high end is rebuilding the core on a solid base while keeping the screens and flows you already worked out. A short review of the code first, usually $2,000 to $6,000, tells you which of the two you need.
Honestly, these tools are great at the first version, and a founder who arrives with a working prototype has already answered the hardest question, which is what the product should do. So this guide covers what building it yourself costs, what an AI-built prototype is usually missing, and what moving it to production takes, so you can decide what to keep and what to pay for.
Vibe codingBuilding software by describing what you want to an AI tool in plain language and accepting the code it writes, usually without reading it line by line.
What it costs to build an app with Lovable, Bolt, Replit or Claude Code
The tools themselves are cheap. Each one sells a monthly plan with a set amount of usage, counted as credits, tokens or messages, and the prices below come from their own pricing pages as of October 5, 2026.
| Tool | Paid plans start at | What the plan gives you |
|---|---|---|
| Lovable | $25 a month (Pro), $50 (Business) | 100 credits a month and up, plus 5 build credits a day, and sync to your own GitHub |
| Bolt | $25 a month (Pro) | 10 million tokens a month, with no daily cap |
| Replit | $20 a month (Core), $100 (Pro) | $20 or $100 of usage on its most capable models each month |
| Base44 | $16 a month (Starter, billed yearly) | 100 message credits a month |
| Claude Code | $20 a month (Claude Pro), Max from $100 | Included in the plan's usage, working in a project on your own computer |
What the AI app builders charge, 2026 · Source: Each tool's own pricing page, checked October 5, 2026. Hosting, the database and paid services the app uses are extra.
The subscription is rarely the real cost, though. Most of it is your time, and the credits you spend going round in circles when the AI fixes one thing and breaks another, which usually starts once the app grows past a few screens. Hosting, the database and any paid services the app connects to come on top, although they are usually small at prototype stage.
So the honest answer to "how much does it cost to build an app with Lovable" is a few hundred dollars and a few weekends for a prototype. The cost that matters comes later, when real people start signing up and paying.
What an AI-built prototype is usually missing
The parts these tools get right are the parts you can see, so the screens, the flows and the basic logic. The parts they tend to miss are the ones nobody sees until something goes wrong, and the biggest is who is allowed to see what.
In May 2025 a security advisory, CVE-2025-48757, showed that apps generated by Lovable were going live with database rules that let anyone read or change tables without signing in, because the access rules on the database were missing or too loose. That could expose names, email addresses and API keys to anyone with a browser, and the fix was to write those rules properly for every table.
The second gap is a safe line between testing and real data. In July 2025, Replit's agent deleted the live database of an app SaaStr's founder was building, after he had told it not to make changes, and then filled it with about 4,000 made-up records. Replit restored the data with its rollback, but it is a good reminder that an AI working directly on live data needs a separate test copy and backups.
After that it is the usual list. Payments that only work when everything goes right, secret keys sitting in code the browser downloads, no handling for errors or empty screens, no tests on the flows that make money, nothing watching the app once it is live, and pages that slow down once the database has real amounts of data in it.
Keep it, fix it or rebuild it?
Most prototypes are worth keeping more than founders expect. If the code is in a normal stack, like React with a Postgres database through Supabase, and the data model makes sense, then fixing it is usually cheaper than starting again, and you keep everything you already tested with users.
A rebuild makes sense when the data model is wrong for where the product is going, when every change breaks something else, or when the app lives inside a platform you cannot take the code out of. Even then the prototype is not wasted, since the screens and flows become the specification, and that is honestly the most expensive part of any first build.
What taking it to production costs
The hours below come from the same work on our own projects at Axtra Studios, so sign-in, database rules, payments and launch, and the dollar column is those same hours priced at what US agencies charge, which Clutch's web development guide puts at $100 to $149 an hour.
| What you get | Time | Hours | At US agency rates |
|---|---|---|---|
| A production review: the code, database rules, sign-in, payments and hosting, with a written fix list | 3 to 5 days | 20 to 40 | $2,000 to $6,000 |
| The prototype you have, fixed and launched | 2 to 4 weeks | 80 to 200 | $8,000 to $30,000 |
| The core rebuilt on a solid base, keeping the screens and flows | 4 to 8 weeks | 200 to 500 | $20,000 to $75,000 |
What taking an AI-built prototype to production costs in the US, 2026 · Source: Time and hours from Axtra Studios projects. Rates from Clutch's web development pricing guide (US agencies, $100 to $149 an hour). Checked October 5, 2026.
The review is the step I would never skip. It is a few days of reading the code, the database rules, the sign-in and the hosting, and it ends with a written list of what to fix, in order, with the hours next to each item, so you know which row you are in before you commit to either.
Where the hours go when you fix a prototype
Here is the middle row broken down, which is keeping the prototype and making it safe to launch. The total comes to 80 to 200 hours, and most of it is the part nobody sees in the demo.
| Line item | Hours | At US agency rates |
|---|---|---|
| Sign-in, permissions and database rules for every table | 20 to 50 | $2,000 to $7,450 |
| Payments and their webhooks, including failures and refunds | 10 to 30 | $1,000 to $4,470 |
| Secret keys out of the browser, and a separate test copy with backups | 10 to 20 | $1,000 to $2,980 |
| Error, empty and loading states | 10 to 30 | $1,000 to $4,470 |
| Tests on the flows that make money | 15 to 35 | $1,500 to $5,215 |
| Logging, monitoring and alerts | 5 to 15 | $500 to $2,235 |
| Launch: domain, email, hosting and handover | 10 to 20 | $1,000 to $2,980 |
Where the hours go when you fix and launch a prototype · Source: Hours from Axtra Studios projects, totalling 80 to 200. Rates from Clutch's web development pricing guide (US agencies, $100 to $149 an hour). Checked October 5, 2026.
Security comes first because it is the one that can hurt you after launch, and payments come second because a payment that fails quietly costs real money. The rest is what makes the app feel finished, so error states, a test copy of the app to try changes on, and something that tells you when it breaks before your customers do.
How long it takes
A review usually takes three to five days. Fixing a prototype and launching it usually takes two to four weeks, and a rebuild that keeps the screens and flows usually takes four to eight weeks, which is faster than a normal MVP launch because the hardest decisions are already made.
We work in two-week sprints with a working version at the end of each one, so you can keep showing the app to users while it is being fixed. If your prototype already has paying users, we can usually do most of the work on a separate copy and move everyone over in one planned release.
Should you just keep prompting?
For some apps, yes. If it is an internal tool for your own team, a demo for investors or a test of an idea with a few friendly users, and there is no personal data or money involved, then I would honestly keep building it yourself and spend nothing on an agency yet.
The line is real customers. The moment strangers sign up, upload their data or pay you, the gaps above stop being theory, and that is the point where a review is worth far more than it costs. Our guide to MVP cost and timeline covers what a first version built properly from the start costs, if you are deciding between the two.
Who owns the code?
You should, and with most of these tools you can. Lovable says you own the code you build with it and can sync it to your own GitHub account, and Claude Code writes the code straight into a project on your own computer. Before anyone works on it, the code should sit in a GitHub account you control, along with the database and hosting accounts.
That matters for the handover too. Whoever takes the app to production should work in your accounts, so the code, the data and the history stay yours whatever happens next, and that is how we work on every web app we build.
Can you keep using the AI tool afterwards?
Usually, yes, and for a lot of founders that is the best setup. Lovable's GitHub sync works both ways, so changes a developer pushes to the main branch come back into Lovable, and you can keep making small changes to the screens yourself while the parts that need care stay with a developer.
The thing to agree on early is who touches what. If you keep prompting changes to the sign-in, the database or the payments after they have been fixed, the AI can quietly undo the work, so we usually agree that those parts go through a developer and everything else stays open to you. Claude Code works the same way in practice, since it edits the same code in the same repository.
What to prepare before you ask for a quote
You will get a much more accurate number if you can give access to the code and say which tool built it, who uses the app today and roughly how many, and whether it takes payments or stores personal data. A short list of what is broken or slow right now is worth more than a long list of new features.
Our production readiness checklist is a plain spreadsheet with the checks above, how to try each one yourself and why it matters. Going through it before a call will really tell you a lot about where your app stands.
When we are not the right fit
If your app is built in a no-code tool like Bubble, there is no code to take over, so the job is a rebuild from the start and our MVP launch work is the better fit. And if you want to keep building it yourself, a review on its own may be all you need, and we will tell you that on the first call.
We are also not the right choice for a quick fix to one bug in an app you are happy with, since a freelancer on the tool's own community will usually do that faster. Where we fit is the middle of this page, so a prototype with real users or real money coming, made safe and launched by one team.
How to get a number for yours
The table will tell you which row you are probably in, and a review will tell you for certain. We can look at your prototype, the tool it was built with and who uses it, and send you a written plan priced in milestones.
If you are adding AI features to the app itself, our guide to AI chatbot and agent cost covers the build and the monthly bill, and if it needs to become a phone app, the guide to mobile app cost in 2026 covers that side. Our guide to web app cost in 2026 covers building the whole thing from scratch, and our AI expertise page has more of the AI work behind this guide.
Prices in this guide come from the Lovable, Bolt, Replit, Base44 and Claude pricing pages, Lovable's GitHub documentation and Clutch's web development pricing guide, all checked on October 5, 2026. The incidents come from Matt Palmer's CVE-2025-48757 advisory and The Register's report on the Replit database, and everything here follows our editorial policy.
