Costs and usage

Your wallet is one balance, but a project spends it in several places at once. Two reports cover it, because they answer different questions.

Costs adds everything up by month: a chart or a table of what each category cost, and whether it is growing. AI Usage lists your AI sessions one by one, for when the question is which requests were expensive rather than what the month came to.

Both are reachable from the Dashboard: the Costs card opens the first, the AI Usage card the second.

The categories

Every cost belongs to one category. Open a category on the Costs report to see what it was made of that month.

AI

Agent turns, Asks and Database queries you ran in this project, priced from the points each one used. The breakdown groups them by the kind of work. The AI Usage report lists every session.

Background AI

AI work Sapilon ran for the project without you asking: looking into an incident, reading a file you uploaded, tidying a task draft. It is kept apart from AI so your own conversations are not made to look more expensive than they were.

Experts

Expert work you accepted a quote for. Every job is listed on its own, even two with the same subject.

Servers

The machines your app runs on: the Build Server while you build, and the Rehearsal and Live Servers once you publish. The breakdown names the server first, then the part of it that cost something and what drove that cost:

• Development machine: the Build Server, charged by the hours it runs. Your organisation’s first 50 hours are free; after that it costs €1 an hour.
• App server and API endpoint: charged by the calls they answer (for example “12,400 calls”), plus the processing time each call used.
• Container server and Virtual machine: charged by how long they run (for example “10 hours of running”).
• Load balancer and Network and disks: the pieces that connect those machines.

Database

Your app’s database on the Rehearsal and Live Servers: the hours its server runs and the space your data takes.

Files

Storage for the files your app keeps, such as uploads, images and generated documents. It is what your project’s storage costs in total, not a charge per upload, so deleting one document will not move it visibly.

Delivery

Getting your app to people’s browsers: fast delivery worldwide, your domain names, and data sent out.

Email

The emails and notifications your app sends.

Security

What protects your app: the web firewall, encryption keys, stored secrets and sign-in.

Monitoring

The logs, metrics and alarms that keep watch over your app.

Other

Cloud services that fit none of the categories above. If something here looks unexpected, tell us: it usually means we have not named that service yet.

Reading it honestly

A category with nothing in it is not shown. A missing Servers line on the Dashboard means nothing has cost anything yet (Build Server hours inside your free 50 add nothing), not that hosting is free. We would rather show you less than assert a €0.00 you might plan around.

Hosting is about a day behind. Cloud costs are reported to us on roughly a 24-hour delay, so the card says which day the hosting figures are current as of. Today’s hosting is never in today’s total, and a project that went live this morning will show nothing under Servers until tomorrow.

Nothing appears before it is real. Costs land here when they are incurred, not when they are forecast. If you want the forward-looking number for going live, that is the estimate in the Go-Live wizard, not this page.

Reading the Costs report

It is monthly, always, because cloud costs are billed and read by month. The bar at the top chooses what you see:

• Server. Narrow to the Build, Rehearsal or Live Server.
• Category. Narrow to one category. Picking one also opens what it is.
• Chart or Table. Two ways to read the figures.
• Month. In the table only: the month it describes. It starts on the current month.

The chart shows each month as a bar, stacked by category, with a breakdown when you hover a month. It shows the last twelve months, or six while your project has less than six months of history. Months you spent nothing in still get a place: a quiet month is worth seeing, not worth hiding.

The table shows one month: a row per category with its cost, its share and how many records it came from, and a total row at the bottom.

Click any category (a row in the table, or a name under the chart) to see what it is and what it was made of (in the month you picked, or across the chart’s months): which server and what on it for infrastructure, the individual jobs behind Experts, or the kinds of work behind AI. Each category links to its section on this page.

Filtering by server

The server filter narrows infrastructure, not the whole bill. Filtering to the Live Server narrows servers, database and the rest, but expert work and your own workspace AI sessions do not belong to a server, so they still count. A filtered total is never smaller than what you actually owe for that server.

On the AI Usage report you can also filter by date, feature, origin and mode.

If a number looks wrong

Report it. Billing questions are free to ask, and a mis-metered cost is our defect to fix, not something to reconcile on your side.

Next: What a turn costs · Wallet and top-ups · Points