Get token usage aggregated into time buckets (15 minutes by default) by user and model.

Returns usage broken down by:

• input_tokens: Total input tokens as the provider reported them
• cached_tokens: Cache read tokens
• cache_write_tokens: Cache write tokens
• new_input_tokens: Input tokens billed at the full input rate
• output_tokens: Total output tokens
• reasoning_tokens: Reasoning output tokens (subset of output_tokens)
• billable_tokens: Every charged token, comparable to a provider console

Defaults to the caller's own data. Admin users (with @quadrillion.io email) may opt into cross-user data by passing user_id (specific user) or all_users=true (everyone). Non-admin requests for other users' data are silently scoped to the caller.

Passing org_id scopes the response to that organization's members and takes precedence over user_id/all_users. It requires platform admin or org admin/owner access (403 otherwise) and includes member identity in buckets.

Workloads read aggregate usage without identity unless explicitly granted platform administration. Admin workloads can also filter by user or org and see account identity.