Monitoring usage
Every API call is metered and billed the same way as the rest of AI HUB, so the console shows the complete picture. There is no reconciliation and no surprise invoice.
How usage is recorded
Each request writes exactly one usage record once the response completes, capturing the model, the prompt and completion tokens, and the cost. The cost is the model's list price applied to those tokens, and it is debited from the prepaid wallet immediately.
- Streamed responses are recorded when the stream finishes, including when the client disconnects mid-stream.
- Failed calls that never reached the model (auth, quota, rate limit) are not charged.
- Pricing follows the Models table. There is no separate API price list.
Watching spend
The console shows API usage alongside the rest of the account activity. Filtering the usage view to API calls shows the per-call model, token counts and the exact amount charged.
The Developer page also shows per-key request counts and recent activity, which indicate the key doing the work.
Keeping spend in check
Two controls bound what a key can spend:
- Wallet balance: the hard limit. At zero, calls stop with a 402 until a top-up is made.
- Per-key monthly cap: an optional ceiling set per key, useful for separating projects or limiting a shared key.
Manage both on the billing page and the Developer page.