Build usage
Sending messages in Lovable to plan, generate, edit, or update your app
This is you, typing. The one everybody expects, and the only one most people budget for.
One credit is not one message. Building is not the only thing that spends them either. Lovable puts both of those in writing, and most people top up before they reach the page that explains why the balance moved.
Lovable defines a credit as the unit it uses to measure and pay for usage across your workspace. The load-bearing word there is usage, and it covers three separate things, only one of which is you typing.
Start with the simple half. In Plan mode, where the model talks through an approach and never touches your code, every message costs 1 credit, flat. That is not where your balance goes.
Build mode is the variable one, and Lovable publishes four worked examples of what it charges. Making a button gray costs 0.50. Removing a footer, 0.90. Adding authentication, 1.20. Building a landing page with images, 2.00.
So the 100 monthly credits on their entry paid plan are not 100 changes. Measured against those published figures they are somewhere between fifty and two hundred, and which end you land on is decided entirely by what you ask for. Ask for whole features and you reach the wrong end of that range by the second week.
None of this is unique to Lovable. It is how every metered AI bill behaves, and the reason the total keeps landing above the estimate is worked through in why an AI bill comes in over the estimate.
One more line worth knowing before it costs you something: a Build mode request you stop partway is charged for the work completed so far. Hitting stop is not a refund.
Here is the part people learn from the usage chart rather than from the docs. Lovable names three kinds of usage, and all three draw down the same credit balance.
Sending messages in Lovable to plan, generate, edit, or update your app
This is you, typing. The one everybody expects, and the only one most people budget for.
Hosting your app and running its built-in backend (Cloud), including database, network, storage, edge functions, and realtime usage in deployed apps
This is the app existing. It runs whether or not you open the editor, and it bills to the same balance.
AI features your deployed app uses, such as calls your app makes to AI models
This is your customers. If what you built asks a model anything on their behalf, that is metered too.
Read together, those three mean a good week costs you credits. Traffic arrives, the app serves it, the meter runs, and the balance you were holding for next month is smaller than it was on Monday. Nothing has gone wrong when that happens. The app is up, so the meter is on.
Then the expiry rules, published and easy to walk past. Monthly plan credits expire two months after they are issued. The daily grant of five build credits resets at midnight UTC and does not roll over. Credits you buy as a top-up last twelve months from purchase.
You came here wanting a total. There is not one, and any page that quotes you a confident figure for what your shop will cost to build is guessing at prompts you have not written yet.
What can be counted is the work. A shop that takes money needs a catalog with variants, stock that moves when somebody buys, a cart, a checkout, an order record, a refund path that puts the stock back, and a place where a human can see all of it. Against Lovable's own published rate, adding authentication runs 1.20 credits. None of the things on that list is a smaller job than adding authentication.
That is the arithmetic to do before you top up again, and it is the one that catches people out. The storefront is the cheap half. Everything in the panel beside this is the other one, and each line of it is a feature rather than a screen.
Describe what you sell and who you sell it to.
Get it built by AI from front to back.
Your shop goes live ready to take its first order today.
If what you are building never takes money, none of this is a problem you have. Lovable's gallery is bigger than ours and its defaults look finished, and on a brief that is mostly about how the thing looks it will reach a showable screen before we do. Paying by the message for that is a fair trade.
The credit model is not a trick either, and it is worth saying so plainly. Metering hosting and model calls is how the cost genuinely behaves, and Lovable publishes its rates instead of burying them, which is more than most. Our billing is simpler because we carry less: your shop runs on your own hosting, so there is no hosting for us to meter.
One thing they have that we have no answer to. Their plans include a grant of five build credits a day at no extra cost, so early iteration on a small idea costs you nothing at all. If you are still at the stage where the whole thing changes daily, that alone is a reason to stay.
If what pushed you here was the shape of the work rather than the shape of the bill, the version of this argument about what actually gets generated is on replacing Lovable for a commerce project.
And if the shortlist still has two names on it, both vendors' own documentation is read side by side on the head to head between Lovable and Bolt, including where each one stops on a shop.
The rate card is only half of the decision, though, because a cheap credit spent on the wrong tool is still wasted. What the product covers, and the sentence in its docs that decides whether it suits a shop at all, is set out in our full review of Lovable.
Our own numbers, including what a credit costs and what a top-up buys, sit in full on the MaShop pricing page.
Built around what AI costs not what your shop earns.
Try MaShop on a side project. No card required.
Includes:
For people running a commerce app full-time.
Includes:
Built for agencies and high-volume teams.
Includes:
It depends what you spend it on, which is the honest answer and also the annoying one. A Plan mode message is 1 credit. Build mode is priced by the work done, and their published examples run from 0.50 for a styling change up to 2.00 for a landing page with images. Hosting your deployed app and any AI calls that app makes draw on the same balance on top of that. Read on their documentation on 4 August 2026.
Because building is not the only thing they pay for. Lovable names Cloud usage, which is hosting your app and running its backend, and AI gateway usage, which is model calls your deployed app makes, as drawing on the same balance. Both of those run while you are asleep.
Partly, and the rules differ by type. The daily grant of five build credits resets at midnight UTC and does not roll over. Monthly plan credits expire two months after they are issued. Credits bought as a top-up last twelve months from purchase. All three were read on their pricing page on 4 August 2026.
No. Their documentation says a stopped Build mode request is charged based on the work completed so far. Pressing stop cuts the remaining work, not the bill for what already ran.
One meter rather than three. A credit pays for the generation you asked for, and the cost is shown before you confirm it. Your shop deploys to your own hosting account and your own database, so a customer browsing it at two in the morning costs you nothing. Bought credits stay valid a year, plan credits reset each cycle, and the bought ones are spent first.
One prompt, and MaShop builds the catalog, the checkout and the back office around what you sell.