For people on Lovable, Bolt.new, Replit or v0
You can build your own. In one day.
One day is all it takes to get off Lovable for good.
You already know how to tell an AI what to build. That stays the whole job, and it's how I build too, as a software engineer. Lovable just charges you by the meter for it. Pick your plan and see what it costs at one flat price.
Only Lovable measured so far.
Spend the same, build more
22×
more building per dollarExtrapolated
Lovable, $100 a month
275
landing pages a month, with every daily bonusPublished
Claude Max 5x, $100 a month
6,200
landing pages a monthExtrapolated
Under the hood it's the same AI. Lovable runs on Claude, in Lovable's own words, and Claude is sold by the token. Your credits pay that token bill, plus whatever Lovable adds on top. Lovable's guidePublished
Switch in one day ↓How that works out
Claude Max 5x at $100 is the Claude plan closest to Lovable's $100 without going over, and in a month it builds 22× as much. Per dollar means each side's monthly pages divided by its price, so the two figures are the same whenever the prices are.Extrapolated
One day is what the escape plan I'm writing is built around: accounts, a domain, and your first site live. I have not timed a cold run yet. When I do, the number goes here, even if it embarrasses me.
About 200 landing pages' worth, or 800 small edits. Then the tank is empty until next month, or until you feed it a top-up.Published
Up to 275 pages or 1,100 edits if you show up for the 5 daily bonus credits every single day, like it's a mobile game.
About 324 landing pages' worth, or 1,370 small edits, every 5-hour window. Then it refills on its own.Extrapolated
There is a weekly cap on top. Hit it and you wait. Nobody sends you a bill.
The month is a realistic one on Claude Max 5x: about $1,625 of API-priced work, a quarter of what the plan allows on paper. Lovable's figure is the generous one: the plan's credits plus all 150 daily bonus credits, which only count if you use 5 every single day.Extrapolated
A landing page cost $0.26 of Claude Code work in my runs.Measured
Credits run dry. A limit just sends you to lunch.
Building as fast as the plan allows, Lovable's pool is dry on day 1.Published
See the whole month, day by day
Owners you can pitch in a month, one preview site each:
Lovable, 400 credits · $100 a month
52
sites, then dryPublished
Claude Max 5x · $100 a month
1,750
sites of roomExtrapolated
33× as many small sites per dollarExtrapolated
Here's the play: build a cheap preview of a local business's new site, show it to the owner, and sell them the real one. Most say no, so every preview you can afford to build is one more owner you get to pitch. Here's how many previews each side has room for in a month, on the plan you picked above.
One site here is one landing page plus 12 follow-up edits.
Adjust the pace
Lovable, 400 credits · $100 a month
52
sites finished in 30 daysPublished
The pool is bone dry on day 1. After that you live on 5 credits a day, about one site every 2 days.Published
Claude Max 5x · $100 a month
1,750
sites finished in 30 daysExtrapolated
No pool to empty. Same pace tomorrow.Extrapolated
Flat out, the plan is the limit: about 1,750 sites, a realistic month of its 5-hour windows under the weekly cap.Extrapolated
Hit the weekly cap and you wait. Nobody bills you more.
33.7×
as many sites in the same monthExtrapolated
Day 1: the pool is empty
PublishedTo match that on Lovable you would need more than the top listed plan, against $100 for Claude Max 5x.
ExtrapolatedSites finished, day by day
Point at a day, tap it, or focus the chart and use the arrow keys, to see its numbers.
Show it day by day
| Day | Lovable credits spent | Lovable sites so far | Own-stack sites so far |
|---|---|---|---|
| 1 | 405.0 | 38 | 58 |
| 2 | 5.0 | 39 | 116 |
| 3 | 5.0 | 39 | 175 |
| 4 | 5.0 | 40 | 233 |
| 5 | 5.0 | 40 | 291 |
| 6 | 5.0 | 41 | 350 |
| 7 | 5.0 | 41 | 408 |
| 8 | 5.0 | 42 | 466 |
| 9 | 5.0 | 42 | 525 |
| 10 | 5.0 | 43 | 583 |
| 11 | 5.0 | 43 | 641 |
| 12 | 5.0 | 44 | 700 |
| 13 | 5.0 | 44 | 758 |
| 14 | 5.0 | 45 | 816 |
| 15 | 5.0 | 45 | 875 |
| 16 | 5.0 | 46 | 933 |
| 17 | 5.0 | 46 | 991 |
| 18 | 5.0 | 47 | 1,050 |
| 19 | 5.0 | 47 | 1,108 |
| 20 | 5.0 | 48 | 1,166 |
| 21 | 5.0 | 48 | 1,225 |
| 22 | 5.0 | 49 | 1,283 |
| 23 | 5.0 | 49 | 1,341 |
| 24 | 5.0 | 50 | 1,400 |
| 25 | 5.0 | 50 | 1,458 |
| 26 | 5.0 | 50 | 1,516 |
| 27 | 5.0 | 51 | 1,575 |
| 28 | 5.0 | 51 | 1,633 |
| 29 | 5.0 | 52 | 1,691 |
| 30 | 5.0 | 52 | 1,750 |
How it works
Take Lovable out of the middle
Now
- Youdescribe what you want
- Lovablecharges you per credit
- Claudebuilds it
- Live siteon Lovable's servers
After you switch
- Youdescribe what you want
- Claudebuilds it, on a flat plan
- Your codein your GitHub
- Live siteon your domain
Lovable builds your app with Claude. Take Lovable out, and you talk to Claude directly, and the code is yours. Lovable's guidePublished
The setup is the one new step, and I walk you through it.
My receipts
In the last month I used $3,164 of AI. I paid $100.
31×of pay-per-token AI for every dollar I paidReceipts
And that's not maxed out. My busiest week was $1,246. A month of weeks like that: about $5,300.Extrapolated
I'm not on a meter. I pay flat, then I abuse it.
Starting out solo, that's the whole trick. Don't let them take advantage of every bad idea or debugging cycle. Flat doesn't care.
The catch: there's a limit, and when I hit it I wait a few hours. These are my receipts. Yours will be different.
Where those numbers come from
A free tool called ccusage reads the usage logs on my laptop and prices every token at Anthropic's pay-per-token rates. I read the 30 days up to 2026-09-24. The plan is Claude Max 5x.
Most of it ran on models that cost more per token than the one I measured for the calculator, so don't line this up against the landing-page numbers.
The limit is a 5-hour window plus a weekly cap. I've hit the window twice. The week, never.
Same prompt, both ways
Same prompt. Look at what each one built.
I sent Lovable and Claude Code the same one-message prompt: a scroll-driven 3D watch page in seven chapters. One try each, no fixes after. The third is mine after I kept at it.
Lovable, one try
56% of the 5 free credits Lovable gives you a dayMeasured
$0.70 (2.80 credits) · 4 minMeasured
A black ring cuts across the watch in every chapter, the parts barely separate, and they don't go back together at the end.
Claude Code, my setup, one try
8% of one five-hour window on Max 5x, 40% on ProExtrapolated
In API dollars
$6.85 of API-priced tokens · 32 min · 34 testsMeasured
All seven chapters as asked, three finishes you can switch between, and no errors in the console.
Claude Code, my setup, built out
Scroll it yourselfHeavier to run: it does more to look like this. Close a few tabs if it stutters.
About $33 a day over 3 days. The busiest day used about 63% of one five-hour window on Max 5xExtrapolated
In API dollars
$99 of API-priced tokens total, to 2026-09-27Measured
Eight layers, a movement you can look into, glowing hands in the dark, and the watch on its strap at the end.
Both run on Claude. What I add is rules: plan first, write tests, check its own work. The escape plan shows you how to set up the basic ones.
Both Claude columns came out of my flat plan, not a meter.Receipts
One run each, on 2026-09-26; the third is one site, costed to that day. Yours will be different.
What you give up
- The visual editor. You type what you want to Claude in a terminal. Same prompts you write now.
- Logins and a database, done for you. You ask Claude for them, the same way you'd ask Lovable.
- Someone else on call. A deploy breaks, it's yours. Paste the error or a screenshot to Claude.
What you own
- Your code. In your GitHub from minute one.
- Your domain. Every project gets a subdomain.
- Extra projects cost $0. No per-project fee, nobody tapping a watch.
- What you set up. You set it up by prompting Claude, so you know what to ask it.
Switch in one day.
In the escape plan, I guide you through Claude to set up the system once. Then you keep building the way you did before. Leave your email for the early price.
One email when the escape plan ships, and that's the only one. One click unsubscribes. I don't sell the list. I barely want to run one.
Show the math
Every figure on this page wears one of four labels. A calculator that hides where its numbers came from is a sales pitch with a slider, and this one is trying not to be that.
- Published
- Read off the vendor's own page, on the date shown.
- Measured
- From my own runs: Lovable's example prompts through Claude Code three times each, and the watch-page prompt once on each side.
- Receipts
- My own usage logs. My numbers, not a promise of yours.
- Extrapolated
- Worked out from one account's data under the assumptions listed below.
Lovable's four example prompts, run through Claude CodeMeasured
| Prompt | Lovable | Claude Code |
|---|---|---|
| Build a landing page with images | $0.50 | $0.26 |
| Make the button gray | $0.13 | $0.06 |
| Remove the footer | $0.23 | $0.05 |
| Add authentication | $0.30 | $0.37 |
Lovable's credit costs are the examples on its credits page, read 2026-09-19, at $0.25 a credit. Mine are API list prices on Sonnet 5 with prompt caching, 3 runs each on 2026-09-19. Every run ended with a passing build; none was tested in a browser. Claude drew its own illustrations where Lovable generates images, so the landing page is not like for like.
What a Claude plan allowsExtrapolated
| Plan | Price | A full week | A realistic month |
|---|---|---|---|
| Claude Pro | $20 | $300 | $325 |
| Claude Max 5x | $100 | $1,500 | $1,625 |
| Claude Max 20x | $200 | $6,000 | $6,500 |
One small site, the unit in the month view
| Step | Lovable credits | Claude Code |
|---|---|---|
| 1 × landing page with images | 2.00 | $0.26 |
| 6 × button edit | 6 × 0.50 | 6 × $0.06 |
| 6 × footer edit | 6 × 0.90 | 6 × $0.05 |
| One site | 10.40Published | $0.93Measured |
At a fixed pace, the own-stack month stays under what the plan allows: one 5-hour window on Pro is about 18 sites and a realistic month about 350, against 150 at the fastest fixed pace. As fast as the plan allows, the month is that realistic one.Extrapolated
Lovable's side assumes every daily bonus credit is collected, which is the generous reading. Days are counted over a 30-day month.
The assumptions carrying it
- One reading on one Max 5x account put its week at about $1,500 of API-priced work. That is the only data point.
- Pro is taken as a fifth of that and Max 20x as four times. Anthropic states those ratios for the 5-hour window; applying them to the week is an assumption.
- A real person uses about a quarter of the theoretical allowance, so the recommendation only counts a quarter.
- Anthropic does not publish these limits, and how they are metered across models is not public.
Receipts: in the last month I used $3,164 of API-priced tokens on a $100 Max 5x plan. The catches are up the page, under my receipts.Receipts
An estimate, not a quote. I'm a guy with a calculator, so check your own Lovable usage and Anthropic's current plan pages before you move anything.