OpenAI Usage API Now Supports API Key Grouping for Token and Cost Reconciliation
OpenAI's August 4, 2026 API changelog introduced API-key filtering and grouping across its usage and cost endpoints, enabling developers to attribute token activity and spending to specific keys. Because the two endpoints return different datasets, a simple inner join risks hiding unmatched or unattributed rows, making a full-outer join on date and API-key ID the safer approach. Both endpoints support daily bucketing and pagination, and consistent grouping dimensions across both sources are required to ensure row-level compatibility. A .NET implementation demonstrates the pattern using synthetic data, flagging each row as matched, usage-only, or cost-only to keep gaps visible rather than concealed. Notably, the approach deliberately avoids estimating costs from token counts, since cost buckets may include line items beyond completion tokens.







