Usage and cost breakdown¶
Track how much you are using Mango Inference and what it costs.
View usage in the console¶
Sign in at https://inference.mangoboost.io. The Dashboard (https://inference.mangoboost.io/dashboard) is the usage summary; Billing (https://inference.mangoboost.io/billing) holds spend and balance.
Which breakdowns the console offers (by model, by key, by day) is not documented here, because the console is the thing that would go stale first. Open it and look; if a dimension you need is missing, that is worth saying in Discord.
Build your own breakdown¶
The console is a view, not an export: there is no cost/usage API today. If you need usage sliced your own way, the durable source is the API response itself. Every completion reports what it consumed:
"usage": {"prompt_tokens": 14, "completion_tokens": 247, "reasoning_tokens": 237, "total_tokens": 261}
Log that alongside the request's own context and you can reconstruct any breakdown you want:
| To break down by | Log alongside usage |
|---|---|
| Model | The model field from the response. |
| API key or environment | Which key your client used. The API does not tell you. |
| Feature or endpoint | Your own call site, tenant, or route. |
| Individual request | x-request-id from the response headers. |
This is worth setting up early rather than after the first surprising bill: it is the only record that survives, it attributes cost to the part of your system that caused it, and it works the same whether or not a usage API ships later.
Watch reasoning_tokens in particular. It is included in completion_tokens,
not added to it, but on a reasoning model it can be most of the generation.
See Reasoning.