OpenAI's Usage and Costs APIs now break spend down by API key
First-party per-key attribution lands in the dashboards and the reporting APIs — bill each customer, meter every agent, catch a runaway key before the invoice.
Copy markdown
What shipped: a per-key cost dimension
OpenAI's Usage and Costs dashboards can now filter and group by API key, and the Usage API and Costs API expose the same API-key dimension for programmatic reporting. Before this, spend came back as one blended total.
Why it matters for you
Run a multi-tenant app? Issue one key per customer and you can bill each exactly. Issue one per agent or project and you can see which is burning budget — instead of guessing from a single lump invoice.
Attribution, not a hard cap
This is reporting granularity, not enforcement: it tells you where the money went, it won't stop a key mid-spend. Pair it with your own budget alerts, or an AI-gateway layer if you need hard per-key ceilings.