How many client sites one account will carry

How many sites a month, and what moves the number

Start with the arithmetic, because it is the only part anyone wants at five o'clock on a Monday. One account carries as many client sites as its credits carry build turns, and nothing anywhere counts the projects. On Studio at $299 a month, 10,000 credits is from 100 build turns to 500, decided entirely by which model each turn runs on. If a job takes twenty turns end to end — one first build and nineteen changes — that is from five client sites a month to twenty-five. The spread is not evasion. It is a fivefold difference in the model you pick, and it is the single biggest lever you have.

What you are delivering at the end of those turns is a standard React and Vite project. Download the whole project as a zip, or push it to your own GitHub or Vercel account. Export to GitHub or Vercel sits on the free tier, not behind a paid plan, so you can see exactly what the artefact looks like before you have paid anything. That is the right order to do it in when the deliverable is going out with your name on it.

Every price on this page is in US dollars. That is what the checkout charges whichever country you invoice your own clients from, so treat the plan as a fixed cost in dollars sitting underneath jobs you price in your own currency. And the last column of the table is the honest one: what moves the number of sites you get through is which model the turn runs on, how many pages each site has, and how many rounds of changes the client asks for. Not the name on the plan.

Note. There is no project limit anywhere on the server, on any plan. The figure in the table is a credit allowance, not a site allowance, and keeping the two separate in your head is most of understanding this page.

What one build turn takes out of the month

A build turn is one instruction and the work that follows it. Building a whole site from a description is a turn. So is moving the contact form above the photographs. The published rates are 100 credits on Opus 5, 60 on Sonnet 5 and 20 on Haiku 4.5 — the same turn at three prices, five times apart end to end.

Divide those into an allowance and you have the range a plan actually buys. It is a range rather than a figure because a single number would have to pick a model on your behalf, and whichever one it picked would be wrong for half the people reading.

One qualification, and it matters if you are about to sign a fixed price. Billing is metered: a hold is taken before the turn and the difference refunded after. The per-model figure above is the hold — the amount reserved before anyone knows what the turn will use — and the charge that settles is priced on what the turn actually used. Plan with the hold, because it is the pessimistic number and the one that will not embarrass you in front of a client, then look at what two or three finished jobs really cost before you quote a fourth at a flat fee.

Nobody has measured a typical client site in turns, and this page is not going to invent an average for you to size a plan against. Count your own. Run three jobs, note what each one took from the first brief to the handover, and you will have a figure that is true for the sites you build rather than for a sample of somebody else's. It will also be a better sales asset than any of this, because it lets you say what a fourth round of changes costs you before you agree to one.

• 300 credits a month: from 3 build turns to 15.
• 20,000 a month: from 20 turns to 100.
• 50,000 a month: from 50 turns to 250.
• 100,000 a month: from 100 turns to 500.
• 200,000 a month: from 200 turns to 1,000.

Note. A turn is priced on what it used rather than on what was reserved for it, so the two ends of the range are the planning figures and not the invoice.

Nothing counts the projects

Worth stating flatly, because most builders charge for it and readers assume it applies here: there is no project limit anywhere on the server, on any plan. Not on Agency, not on Trade, and not on the free tier either.

That last part is precisely why it is not a selling point. A limit that no tier has is not something $299 a month is buying you, and presenting it as a paid feature was an error that has already been corrected once. So take it as a fact about how the thing is built rather than as generosity. What the money buys is throughput, not permission to take on a fourth client.

What is counted is short, and all three are enforced rather than advertised: the credits granted each month, how many builds run at the same time, and how many builds you may start in an hour and in a day. Three walls. The number of projects sitting in the account is not one of them, and neither is who those projects belong to.

• Counted: credits granted each month, and spent per turn.
• Counted: builds running at the same moment.
• Counted: builds started within an hour, and within a day.
• Not counted: how many projects sit in the account.
• Not counted: how many different businesses those projects are for.

What the plan buys, and what the client buys

These are two different purchases, and running them together is what makes the pricing feel strange when you first look at it. Your plan buys throughput: turns in a month, builds at once, a ceiling per hour. The client buys an artefact and the judgement that shaped it. One is a standing cost of yours and the other is a deliverable of theirs, and the exchange rate between them is no more the client's business than the rent on your desk is.

Which is the argument this page rests on. Using an AI builder on paid client work is a delivery choice rather than a compromise, because the client ends up owning downloadable code either way.

Whether the components were typed by hand across three days or generated across twenty turns, what lands at the end is the same class of thing: a React and Vite project the client can hold, in a zip or in a repository in an account with their name on it. Nothing they own is worse for how it was made. A deliverable that is as good and arrives sooner is delivery, and the honest word for it is not an admission.

The arithmetic is what makes that a business rather than a trick. A month on Studio is $299, and it does not scale with the number of clients — it scales with how many turns you run, which is a number you decide by how you scope and how many rounds you agree to. Ten clients whose sites are finished and quiet cost the same as none. That is a very different cost curve from the one a studio built around billable build hours is carrying.

One part genuinely does not survive, and naming it is what makes the rest believable. If what you sold was the labour of assembling a template, that labour is worth less this year than it was last year, and no tier fixes it. Most people doing this work were never selling only that. They priced as though they were, because hours are easy to explain on an invoice and judgement is not.

Note. The test: if a client asked what they got for the fee, could you answer without mentioning how long it took? If not, the problem is the brief you accepted, not the tool you built it in.

The number that bites on a busy Thursday

Credits run out at the end of a month, which you can see coming. Concurrency bites at three in the afternoon, which you cannot, and it happens while somebody is waiting on the other end of a call.

A build turn holds a connection open for up to three minutes, so the limit that maps to real capacity is not how many builds you may run in a month but how many may run at the same time. The free tier allows one. Trade allows 2, Growth 3, Studio 5, Agency 8.

Working on one client at a time, this never surfaces. On a Thursday when two clients have both sent their changes back and a third site is being built from scratch, it decides whether you work in parallel or in a queue. For a lot of client work that is the real difference between one tier and the next, and it is not the number people compare on the pricing card.

There are hourly and daily ceilings sitting behind it as well. They exist so that a script cannot empty an account in an afternoon, and a person working with their hands on a keyboard is not going to meet them: 60 builds an hour on Trade, 150 on Growth, 400 on Studio, 1,000 on Agency. Worth knowing they exist. Not worth planning around.

Note. If you build in front of clients while you are on a call with them, concurrency is the tier decision rather than the credit figure. Two calls that overlap, plus a build you left running from the morning, is three at once.

When a month runs long

Some months a client comes back four times when the quote said two, and the allowance goes before the renewal date does. The fix is a one-off pack rather than a tier: 400 credits for $5, 1,200 for $15, 2,500 for $30, or 6,000 for $70.

The part that decides how you use them is where they sit. Top-up credits sit in a second bucket that never expires and is spent after the monthly allowance. So buying a pack in a heavy month does not hand you a balance to chase before it disappears — the monthly allowance is the thing that expires at renewal, and the pack is the thing that does not. Spending takes the allowance first, which means the credits that were going to vanish anyway go first and the ones you paid for outright wait behind them.

That makes the sensible move at the end of a long month obvious. Buy the smallest pack that gets the job over the line, rather than moving up a tier because of one client with an unusual appetite for revisions. A tier is a standing cost every month afterwards, and a single bad month is not evidence about the next one. If three months in a row run long, that is evidence, and then the tier is the right answer.

The other half of the same discipline is on the quote rather than in the account. Write the revision ceiling into the quote as a number, and a price for the round after it. A month that runs long is almost never a plan that was too small — it is a quote that said reasonable amendments and left the ceiling to be discovered.

• 400 credits for $5 — one job that overran.
• 12,000 for $15 — a full extra site's worth of turns, at the dearest model.
• 25,000 for $30 — more than a whole month of Trade, bought once.
• 60,000 for $70 — cheaper than moving from Growth to Agency for a single busy month.

Note. Move up a tier when concurrency is what is hurting you, not when credits are. Credits have a one-off fix that never expires. Builds queueing behind each other do not.

Who should not pay for the bigger tiers

The tier names invite an assumption, so here is what they do not come with. There are no seats, and no way to add a colleague or a client to your account. There is no client login and no read-only view of a build you can hand somebody. There is no per-client billing: your card pays for every build on the account whoever the work was for, and there is no route that puts a project on a client's card. There is no dashboard that presents a set of projects as a book of clients.

Studio is $299 a month and Agency starts at $399, both written for people who build sites for other people. What those names actually contain is credits, how many builds run at once, how many an hour, and the support you get. That is the complete list. Nothing in either tier changes anything the client sees at the end, and neither of them is a reseller arrangement behind a different name.

So do not move up expecting client management. Move up when you have counted your turns and run out of them, or when builds are queueing behind each other on a day you needed three at once. Everything else on that list is a spreadsheet and a folder of invoices, the same as it was before.

There is a job-level version of the same honesty, and it is the more expensive one to get wrong. If the client needs a content system their staff will edit weekly, an integration with the booking or stock software they already run, or a design their brand team signs off page by page, quote for the hand build and stop reading here. Those are real requirements. Taking the job anyway is how a sound delivery decision turns into an apology in week five, and no tier on the ladder above changes that.

• No seats. One account, one person paying for it.
• No client login, and no read-only view of a build.
• No per-client billing. Every turn on the account is on your card.
• No dashboard that treats a set of projects as a book of clients.
• Nothing in the tier names alters what the client receives at handover.

Common questions

Does anything in what I hand over reveal that it was built by AI?

Not the layout, and not the code, which is an ordinary React and Vite project any developer can open and read. What gives a site away is generic wording and stock photography, and both of those are your decisions rather than the tool's. Put the client's own words and the client's own photographs on it and the question stops being asked. Read the README and the project files before you push them anywhere, the same as you would read anything going out under your name.

What leaves my account when the job is finished?

The project. Download the whole project as a zip, or push it to your own GitHub or Vercel account. Export to GitHub or Vercel sits on the free tier, not behind a paid plan, so nothing about the handover depends on which tier you happen to be on that month. Push it into an account in the client's name if you want ownership settled on the day rather than argued about a year later.

If I drop from Studio back to Trade, what happens to the sites already on the account?

There is no project limit on any plan, so there is no cap for a downgrade to push you over. What changes is the throughput: a smaller monthly allowance, two builds at once instead of five, a lower hourly ceiling. Anything a client already holds — a zip on their machine, a repository in their own account — does not depend on your plan at all, which is the reason to do the handover at the end of the job rather than whenever somebody asks.

The plan costs me $299 and I am charging thousands. Is that defensible?

You are not billing for the plan any more than a builder bills for the van. You are billing for the brief, the decision about what the business actually sells, words a customer recognises, the revisions, the handover and someone to ring in March when the prices change. If the honest answer to what the client got is only the two minutes the build took, the fee is wrong. If it is that list, the cost of your tools is not the client's business.

What if the client comes back in two years and I no longer have a plan?

They hold the source code, so any competent developer can pick it up — which is more than can be said for a site trapped in a builder nobody can export from. Your plan is your production capacity, not their dependency, and lapsing it costs you the ability to run more turns rather than costing them the site. Say this in the quote instead of saving it for an awkward call. It is one of the strongest lines you have and it costs nothing to give away.

Do my clients need accounts of their own, or can I add them to mine?

You cannot add them to yours. There are no seats and no client logins, so there is no halfway arrangement where they watch a build or edit a page under your billing. They need an account of their own only if you want them changing the site by describing changes rather than through you. The free tier grants 3,000 credits a month, a full site generated and browser preview, which is enough for them to find out whether they want that. Paid plans start at $20 a month, or $16 a month paid yearly.