Credits, tokens and what it really costs

What a credit actually is

A credit is a unit of work done by the model on your behalf. It is not a minute, not a user and not a licence. Describing a change and having it built spends credits; the software running afterwards does not.

Underneath, models are billed by the token, which is roughly a word fragment. Credits are that, rounded into something a business can plan with, so you are not reading a bill denominated in fractions of a cent per paragraph.

The practical consequence is the one people find surprising: the meter runs while you are building and stops when you stop. A system that ten colleagues use every day for a year, with no changes, costs the monthly draw and nothing else.

What spends credits and what does not

Building and changing spends them. Using does not. AI features inside your own running system also spend them, because each of those is a call to a model.

That third and fourth row are the reason this pricing looks unusual next to per-seat software. Adding the tenth colleague costs nothing, because nothing about them is metered.

The monthly draw, and why it is credits rather than euro

Keeping your systems live draws 750 credits a month from the same balance: hosting in Frankfurt, the database, continuous backups, platform updates and support. It is one draw for the whole account however many systems it holds.

It is taken in credits deliberately. There is no subscription, nothing renews, and no card is kept on file, so there is no monthly charge to cancel and no dunning if a card expires. An account that stops buying credits simply stops drawing, and if the balance empties there are fourteen days of grace before anything is switched off.

An account with no system running draws nothing at all. Building something and deciding against it does not leave you paying for it.

What the packs cost

Sizes bought when you need one, never expiring: 5,000 for €77, 15,000 for €227, 25,000 for €377, 40,000 for €597, 60,000 for €877 and 100,000 for €1,447. Prices exclude VAT, which is added at checkout, and every purchase gets a proper invoice.

None of them carries a deadline: these are the standing prices, and the rate per credit falls as the pack grows rather than as a limited offer.

New accounts also get starter credits: 500 at registration, another 2,500 the first time you open the studio, and a top-up if they drop below 500 while you build, up to 5,000 over 7 days, with no card. Whatever is unspent when the week ends goes back, which is the opposite of bought credits, and it is worth planning around: the week runs from registration, so start the same day you register.

What a real first system tends to cost

A small internal system - one trade, a few record types, two roles, some reminders - commonly lands inside a single 5,000-credit pack, with changes over the following weeks inside the same pack or the next one up.

The spread is wide and it is honest to say so. Something with one kind of record and two screens can be done in a fraction of a pack. Something covering a whole operation with six record types, imports, roles and reports is a 15,000-credit job or more, and a company replacing several spreadsheets at once should plan for that rather than be surprised by it.

The variable is not the size of your company. It is how settled the process is. The same system described by someone who knows exactly what they want costs a fraction of what it costs when the requirements are being discovered mid-build.

The four habits that make it cost more

Vague descriptions, rebuilding instead of changing, importing a mess, and building everything at once. Each is avoidable and each is common.

• Vague description: every ambiguity becomes a round trip, and each round trip is credits. Ten minutes of writing saves more than any other economy here
• Rebuilding instead of changing: asking for the whole thing again to fix one field is the most expensive way to move a button
• Importing a mess: fifteen years of inconsistent records will consume more than the system itself. Clean the spreadsheet first, or import only what is live
• Building everything at once: the fifth screen nobody has asked for is paid for now and evaluated never

The part we cannot tell you in advance

Nobody can quote you exactly before you start, and any supplier who does is quoting your uncertainty as well as their work.

What we can do is make the uncertainty cheap to resolve. The starter credits exist so the first version is built before any money moves, and the reason the pricing page is a shelf of packs rather than a contract is that you stop whenever you have what you need.

If it turns out your project is one where every hour has to be predicted in advance, the fixed-fee launch service is the honest answer, and we will say so rather than sell you credits for it.

QUESTIONS