What does a vibe-coded app really cost? Credits, APIs, hosting
Read in : Français

Contents
- What are the cost lines of a vibe-coded app?
- How much do Lovable, Bolt and Replit cost?
- Why credits disappear faster than you expect
- How much do Cursor and Claude Code cost?
- The AI API inside your app: the cost that scales with users
- Worked example: 1,000 users, 20 requests each
- How much do the database and hosting cost?
- What about payments?
- Three sample budgets
- A prompt to estimate your app's running costs
- No-surprises checklist
- Frequently asked questions
- Frequently asked questions
- Sources
The cost of a vibe-coded app comes down to four lines: the tool that writes the code, the AI your app calls, the database and hosting, and payment fees. A prototype can run for $0. A small SaaS with built-in AI usually lands between $70 and $100 a month, before payment fees. Here is the breakdown, line by line, using prices checked on 21 November 2026 on each official pricing page.
What are the cost lines of a vibe-coded app?
A vibe-coded app has four cost lines, and only the first one is visible when you sign up.
| Cost line | Examples | How it is billed | The catch |
|---|---|---|---|
| Builder tool | Lovable, Bolt, Replit, Cursor, Claude Code | Monthly plan plus credits or tokens | Credits drain fast while debugging |
| AI inside your app | Claude API, OpenAI API | Per million tokens, on every request | The bill grows with your users |
| Data and hosting | Supabase, Vercel, Netlify, shared hosting | Free up to a threshold, then plan plus usage | Free projects that pause automatically |
| Payments | Stripe | Percentage plus fixed fee per transaction | Forgotten when setting your price |
If you are just starting, read the pre-launch security checklist first: one leaked API key can cost more than everything else combined.
How much do Lovable, Bolt and Replit cost?
The three big app builders all start at around $20 to $25 a month for the first paid plan, but they do not measure the same thing.
| Tool | Free | First paid plan | Unit |
|---|---|---|---|
| Lovable | 5 credits a day, capped at 30 a month | Pro: $25/month, 100 monthly credits plus 5 a day | Credits per message |
| Bolt | 1M tokens a month, 300K a day | Pro: $25/month, from 10M tokens, unused tokens roll over | Tokens |
| Replit | No separate free plan described on the pricing page | Core: $20/month ($18 billed annually), including $20 toward the most capable models | Effort-based billing |
Lovable says monthly credits roll over for one month on monthly plans and that top-up credits last 12 months. Bolt says Pro is the cheapest plan without Bolt branding on published sites and with custom domains. Replit's Pro plan starts at $100 a month.
To pick between the three for your project, see Lovable vs Bolt vs Replit.
Why credits disappear faster than you expect
A message does not have a fixed price: it depends on how much work the AI does. Lovable's documentation gives examples: "Make the button gray" costs 0.50 credits, "Add authentication" 1.20 credits, "Build a landing page with images" 2 credits. Plan mode costs 1 credit per message.
Replit works the same way with effort-based billing: a small fix costs less than a complex feature, and each request produces a billed checkpoint. Replit lets you set usage alerts and hard budget limits on the Billing page in Settings.
The real money pit is the debugging loop. When the AI goes in circles on a bug, every attempt burns credits for nothing. Three habits keep it in check:
- Write a spec before you prompt, so you do not rebuild the same page three times.
- Switch to Plan (or chat) mode to diagnose, then Build mode to fix.
- Roll back to a version that worked instead of stacking patches. That is exactly what Git for vibe coders is for.
How much do Cursor and Claude Code cost?
Coding agents such as Cursor and Claude Code charge a subscription with a usage allowance rather than showing a price per message.
| Tool | Individual plans | What is included |
|---|---|---|
| Cursor | Free $0, Plus $8, Premium $30, Super $100, Ultra $200 a month | Plus: Grok models only. Premium: 4x Plus usage, third-party models billed at API list price. Allowance refreshes weekly |
| Claude Code | Through a Claude plan: Pro $20/month ($17 billed annually), Max from $100/month | Claude Code included in Pro and Max, not in Free |
These tools ask for a little more technical comfort, but they do not lock you into their hosting. Cursor vs Claude Code vs Codex covers the differences.
The AI API inside your app: the cost that scales with users
If your app calls an AI model (summaries, a chatbot, text generation), you pay for every request per million tokens, and that bill tracks your user count directly. A token is roughly a chunk of a word. Text you send (input) and text the model writes (output) are priced differently.
| Model | Input (per million tokens) | Output (per million tokens) |
|---|---|---|
| Claude Haiku 5.5 (prompts up to 100,000 tokens) | $0.10 | $0.50 |
| Claude Sonnet 5.5 | $2 | $10 |
| Claude Opus 5.5 | $4 | $20 |
| gpt-6-luna | $0.10 | $0.50 |
| gpt-6.2-sol | $2 | $10 |
Prices checked on 21 November 2026 on Anthropic's and OpenAI's official pricing pages.
Worked example: 1,000 users, 20 requests each
Take an app that summarises meeting notes. Each request sends about 2,000 tokens and gets 500 back. With 1,000 active users making 20 requests a month, that is 20,000 requests.
- With Claude Haiku 5.5: 2,000 × $0.10 / 1M + 500 × $0.50 / 1M = $0.00045 per request, so $9 a month.
- With Claude Sonnet 5.5: 2,000 × $2 / 1M + 500 × $10 / 1M = $0.009 per request, so $180 a month.
Same app, same usage, 20 times the bill. For a simple task, a small model is often enough. Anthropic also notes that models from Claude 4.7 onward use a tokenizer that produces about 30% more tokens for the same text, so leave some headroom.
Above all, keep the API key on the server. A key in the front end is your credit card left on the counter, as explained in why API keys must never live in your front end.
How much do the database and hosting cost?
For a first project, the database and hosting can stay free, as long as you know the limits.
| Service | Free tier | First paid plan |
|---|---|---|
| Supabase | 500 MB database per project, 50,000 monthly active users, 2 active projects, paused after 1 week of inactivity | Pro: from $25/month, $10 compute credits included, never paused |
| Vercel | Hobby: $0, 100 GB data transfer, 1M function invocations a month | Pro: $20 per developer seat per month, with $20 of usage credit |
| Netlify | 300 credits | Personal: $9/month (1,000 credits), Pro: $20/month (3,000 credits) |
Supabase pausing a free project after a week without activity is fine for a prototype, not for an app customers rely on. On Vercel, if you exceed a Hobby limit, the docs say you usually have to wait 30 days before using that feature again.
For where to publish depending on your app, read Vercel, Netlify or shared hosting. For the database, make sure you turn on Supabase RLS before opening sign-ups.
What about payments?
Stripe has no monthly fee on standard pricing, but takes a cut of every payment. In the US, domestic cards cost 2.9% + 30¢ per successful transaction, with 1.5% extra for international cards. In the UK, standard UK cards cost 1.5% + 20p, EEA cards 2.5% + 20p and international cards 3.15% + 20p. Subscriptions run through Stripe Billing add 0.7% of volume.
On a $9 subscription paid with a US card, that is 56¢ in card fees plus about 6¢ for Billing: roughly 7% of the price. The right Stripe setup for a vibe-coded app walks through the integration.
Three sample budgets
Here are three scenarios built from the prices above. They are estimates, not real invoices: your usage will move the numbers.
| Scenario | Stack | Estimated monthly cost |
|---|---|---|
| Prototype to test an idea | Lovable or Bolt free, Supabase free, Vercel Hobby | $0 |
| Side project with a few users | Lovable Pro ($25), Supabase free, a domain | about $25 plus the domain |
| Small SaaS with AI and subscriptions | Cursor Premium ($30) or Claude Pro ($20), Supabase Pro ($25), Vercel Pro ($20), Claude Haiku 5.5 API (about $9 for 20,000 requests) | about $75 to $85 plus Stripe fees |
The SaaS scenario assumes you moved from an all-in-one builder to a coding agent. If you stay on Lovable, hosting and the database (Lovable Cloud) draw on your credits, starting with a monthly grant of 20 Cloud credits on Free, Pro and Business. The guide to building a SaaS with vibe coding covers the road to the first paying customer.
A prompt to estimate your app's running costs
Paste this into your tool before it starts building. It makes the AI spell out the cost lines instead of letting you discover them on the invoice.
Before writing any code, list every external service this app will use
(database, hosting, AI API, email, payments). For each one: what is
billed (requests, storage, tokens, seats), the free tier threshold and
what happens when it is exceeded. Estimate AI API calls per user per
month and suggest the cheapest model that does the job.
Do not quote prices from memory: name the pricing page to check instead.
The last line matters: prices change often, and AI models happily quote outdated ones.
No-surprises checklist
- Turn on spending alerts and caps wherever they exist (Replit, Vercel Pro, your AI API console).
- Put a per-user usage limit on any feature that calls the AI.
- Start with the smallest model and only move up if quality is not good enough.
- Make sure no production project runs on a plan that pauses.
- Build payment fees into your price.
- Re-check pricing every quarter: plans change often, as the free vibe coding guide shows.
Frequently asked questions
Frequently asked questions
Can you build a vibe-coded app without paying anything?
Yes, for a prototype. Lovable gives 5 credits a day (30 a month at most), Bolt gives 1 million tokens a month, and both Supabase and Vercel have free tiers. The limits arrive quickly: few iterations per day, the tool's branding on your site, and Supabase projects paused after a week of inactivity. Enough to validate an idea, not to host paying customers.
Why is my Lovable or Replit credit usage so high?
Because each message is priced on the work done. A simple tweak costs a fraction of a credit, a full feature costs several. Debugging loops, where the AI keeps retrying the same fix, burn a lot for little. Plan before you build, diagnose in chat mode, and roll back to a stable version rather than piling up patches.
How much does an AI API cost for 1,000 users?
It depends on the model and request length. For 20,000 requests sending 2,000 tokens and receiving 500, expect about $9 a month with Claude Haiku 5.5 and about $180 with Claude Sonnet 5.5, based on prices checked on 21 November 2026. Track real usage in your API console from the first users onward.
Should I pay for Lovable or Cursor?
Lovable ($25 a month on Pro) includes hosting and suits you if you never want to touch code. Cursor ($8 to $200 a month depending on the plan) is an editor: you handle hosting and the database yourself, but you are not tied to one platform. Many people prototype on Lovable, then move to a coding agent as the app grows.
Do I need a security budget?
Not necessarily a line item, but you need time. Lovable offers free security scans and Supabase has a built-in Security Advisor. The real cost of security is the incident: a stolen API key can rack up hundreds of dollars of calls in hours. Cap your API spend before launch.
Sources
- Pricing — Lovable
- Plans and credits — Lovable Documentation
- Pricing — Bolt
- Pricing — Replit
- AI billing — Replit Docs
- Pricing — Cursor
- Plans & Pricing — Claude
- Pricing — Claude API Docs (Anthropic)
- API pricing — OpenAI Developers
- Pricing — Supabase
- Pricing — Vercel
- Vercel Hobby Plan — Vercel Docs
- Pricing — Netlify
- Pricing — Stripe (US)
- Pricing — Stripe (UK)

