Coroid bills two separate things, and keeping them separate is what makes the cost predictable.
Capacity — agent slots, charged per slot per month. A standing commitment that sets how much work runs at once.
Model usage — the tokens your work consumes, either prepaid as credits or billed directly by your own provider. Variable with what you actually run.
Coroid applies no markup to provider token rates either way.
Why separate
Bundling them would mean either capping model usage inside a capacity price — so heavy users subsidise light ones — or metering everything, so nobody can budget.
Splitting them means the predictable part stays predictable and the variable part stays visible.
Where things live
| Task | Where |
|---|---|
| Change plan | Settings → Billing → Plans & billing |
| Add or remove slots | Settings → Billing → Capacity & credits |
| Top up credits | Settings → Billing → Capacity & credits |
| See what you are using | Settings → Billing → Usage |
| Attribute spend | Settings → Billing → Detailed reports |
| Invoices | Settings → Billing → Billing history |
| Bring your own keys | Settings → Provider keys |
The two questions people ask
"Why is my work slow?" Almost never billing. A task's duration is set by the size of the change and your test suite. Slots control how many tasks run at once. See Capacity and agent slots.
"Why did spend go up?" In order of likelihood: rework rose, a capable profile got assigned to a high-volume agent, or tasks got broader. See Usage and reports.