/enterprise/orgs/{org_id}/tokens) answers “who or what is consuming tokens, and how efficiently.”
What you’ll see
- Tokens by group — top 10 buckets by total tokens, grouped by project, team, user, or model. Each row shows a usage-source badge (
provider= provider-reported counts,modeled= estimated,mixed) and a proportional bar. - Efficiency vs Org Median table — Tokens/Session, Tokens/Line Changed, and Cache-Hit Ratio per group, with an “Org median” row for comparison. Above-median token-per-session values are highlighted.
What you can do
Switch granularity (day/hour) and group-by (project/team/user/model); filter by date range.Measured context reduction
Enterprise administrators also see Measured context reduction for the selected date range. It breaks down estimated tokens saved by method, including Read cache, Search shaping, Exec compaction, Loop breaker, Hash dedupe, and Test output filter. Each row separates enforced reductions from shadow-mode potential. Shadow mode measures an opportunity while leaving the agent output unchanged.Who sees it
Token consumption and efficiency are visible to portal roles with access. The measured context-reduction breakdown is administrator-only because it aggregates organization-wide telemetry.The org median shown here is computed client-side over the currently displayed buckets, not a backend-computed org-wide figure.