Usage & limits
Every request through the gateway is metered: model, tokens in and out, cost, key, tags, outcome. You read it in the dashboard, with no provider dashboards to reconcile. There is no usage API; the dashboard and its CSV export are the source.
In API responses
Section titled “In API responses”Responses carry the standard OpenAI usage object with token counts: prompt_tokens,
completion_tokens and total_tokens. When streaming, set stream_options with include_usage
to receive it in the last chunk (see Making requests).
Cost is never included in a response.
In the dashboard
Section titled “In the dashboard”- Logs: one row per request, with date, model, tokens in and out, cost, key, tags, finish reason and status. Filter by key or tag, open a row for the detail.
- Statistics: totals for the period. Statistics › Explore breaks spend down by dimension, including tag; Statistics › Trends shows it over time.
- Workspaces: each workspace has its own Usage and Logs pages, limited to its keys.
- Limits: the requests-per-minute and tokens-per-minute limits of your account, and your plan’s caps.
- Models: the catalog your account can call.
- Gateway status: the health of your dedicated gateway.
Export
Section titled “Export”Logs can be exported as CSV from the Logs page. The file, inference-logs.csv, has one row per
request with these columns: Date, Model, Tokens in, Tokens out, Cost (USD), Type, Speed (tok/s),
Key, Tags, Finish, Status. Cost keeps its full precision; Type says whether the request was paid
from prepaid funds or the credit line; Tags are separated by ;. An export contains at most
50,000 rows, within the retention window of your plan.
Retention
Section titled “Retention”Request history is kept for a period that depends on your plan:
| Plan | Logs kept for |
|---|---|
| Starter | 15 days |
| Team | 30 days |
| Scale | 90 days |
| Enterprise | No limit |
Export before a period ages out if you need it for your own records.
Limits at a glance
Section titled “Limits at a glance”| Limit | Scope | Where it is set | When exceeded |
|---|---|---|---|
| Requests and tokens per minute | Account, shared by all keys | Your plan; shown on the Limits page | 429, not billed. See Rate limits. |
| Key budget | One key | API keys page | 429 for that key until its 30-day window resets |
| Prepaid balance | Account | Billing page (top-ups) | 401, all keys blocked until a top-up is paid |
| Credit line limit | Account, per calendar month | Our team | 401, all keys blocked until a top-up is paid |
| Active keys | Account | Your plan: 25 on Starter, 100 on Team, 500 on Scale, no cap on Enterprise | Key creation refused |
See Credits & budgets for how the balance and the credit line work.