What a credit actually buys — August 2026

TL;DR — In the AI app-builder category a “credit” is not a unit of work. It is a unit of billing, and every vendor defines it differently. Below: what one credit buys at Base44, Lovable, Bolt, v0 and Hercules as of August 14, 2026, worked out from their published prices; the three multipliers that turn a $25 sticker into a real bill; and the model we chose instead. Every number carries a numbered source at the end.

The unit is the business model

The category has settled on one revenue model: metered consumption. You buy a tier, the tier includes an allowance, and the allowance drains while you build — and, at several vendors, while your finished app runs.

That is not a moral failing. A metered model is a defensible way to fund a fast product, and every tool below is genuinely good at something. But it changes what a prompt is. On a meter, every message is a small purchasing decision, and the person making it is usually not the person who approves the invoice.

So this post does the arithmetic. Not the sticker prices — the unit prices, and what they multiply by.

Base44: two pools, one ladder

Where Base44 is strong: one of the quickest starts in the category, with auth, database and hosting bundled, and a tier ladder that publishes exact credit counts.¹

The mechanic is two separate credit pools. Message credits are spent while you build — every AI message, including the ones that fix the last one. Integration credits are spent when your live app runs: LLM calls, emails, file uploads, automations. Your app’s users consume your credits.^1,2

The unit price falls out of the Builder tier: $40 a month on annual billing for 250 message credits, so $0.16 per builder message.¹ There are no à-la-carte top-ups; when the pool is empty you move up a tier.² A six-message debugging loop is about a dollar. The number that surprises teams later is the second pool: the bill keeps ticking after launch, in proportion to how much the app gets used.

Lovable: variable credits and a second bill

Where Lovable is strong: a large community and template ecosystem, fast and attractive output, GitHub sync on every plan so you own the code, and paid credits that roll over.³

The unit price from the Pro tier is $25 for 100 credits, so $0.25 per credit.³ But a credit is not a prompt. Lovable’s own examples run 0.5 to 1.7 credits per prompt, and complex changes burn two, three or more.^3,4 Third-party analyses put a working app at $90 to $130 or more a month against the $25 sticker.^4,5

Then there is the second bill. Cloud and AI features your deployed app calls at runtime are metered separately, on usage.³ Two meters, one of which runs after you stop building.

Bolt: tokens that grow with your codebase

Where Bolt is strong: raw speed from prompt to a running full-stack app in the browser, real hosting and database through Bolt Cloud, tokens that roll over, and a May 2026 bump from 10M to 13M tokens on Pro at the same price.^6,7

The unit price: Pro is $25 for 13M tokens, so about $1.92 per million tokens.⁶ Teams pay $30 per member.⁶ The mechanic that matters is not the rate but the slope. Every edit re-sends context, so the token cost of an edit rises with the size of the codebase.⁷ Mature projects get progressively more expensive to touch — the meter charges more precisely because your project succeeded.

v0: per seat, then credits

Where v0 is strong: polished UI output that other tools get compared to, Vercel’s enterprise posture behind it, and real developer mindshare.⁸

The bill has two multipliers. Team is $30 per user per month and includes $30 of credits per user, pooled.⁸ Each message then spends credits by model tier — Mini, Pro or Max — and by tokens.⁹ Seats times credits. Hosting for the deployed app is Vercel usage, billed by Vercel, and is not part of the v0 line at all.⁸

Hercules: credits that bill themselves

Where Hercules is strong: a clear business-buyer bundle with commerce and payments, an MCP integration that lets you drive it from Claude Code, Cursor or ChatGPT, and a strong support reputation.^10,11

The unit price from Pro: $25 for about 75 AI credits, so about $0.33 per credit.^10,12 Every agent interaction is metered, including sessions driven over MCP from your own AI tool — so you pay your AI subscription and the credit.¹¹

Credit auto-billing can move you into the next tier automatically so a running agent does not stall, and extra credits expire in 28 days.¹¹ The free tier is personal and non-commercial; commercial use requires a paid plan.¹³

The three multipliers

Put the five side by side and the tiers priced out above run from $25 to $40 a month.^1,3,6,8,10 Nobody buys a month. A team buys a year, and three multipliers sit between the sticker and the invoice.

Builders times prompts. A vague prompt costs the same as a precise one. So does the fix-it prompt after it. Debugging is where credits go, and debugging is most of building.

Runtime. At Base44 and Lovable the meter keeps running when the app is live; Hercules meters hosting through Cloud Credits past a free allowance and spends AI credits on AI features inside the app.^1,3,11 The successful app is the expensive one.

The tier bump. Base44 sells no top-ups; you climb. Hercules climbs for you. Either way the price you agreed to is the price on a quiet month.

The pricing calculator takes your team, your prompt volume and your months and prices a realistic year on each vendor’s published mechanics. Every assumption it makes is printed next to the number, and every figure carries the same dated sources as this post.

What we did instead

CodeMonster is priced as one organization license. It covers every builder and every app. No tokens. No credits. No meter. You bring the AI subscription you already pay for — the Claude or ChatGPT plan you already hold — and we never meter it, mark it up or resell it.

Your live apps run on managed cloud infrastructure we operate, built primarily on AWS, and nothing is metered per action there either. The trade is that platform-built apps run on the platform; we spell that out plainly, because it is the honest half of the sentence.

This category reprices often. Each source below carries the date we checked it, and we treat any figure older than a quarter as unverified. If one of the numbers above has moved, the date will tell you before we do.