Skip to main content

Organization spend

The Organization screen is the spend overview: what the organization has spent on model calls, how that compares with the preceding period, and which groups and users are approaching or exceeding their limits.

Reading the summary​

The top of the screen shows total spend for the last 30 days, updated daily, with a chart of the daily trend. Alongside it is a short-term comparison: spend over the last seven days against the seven before it, with the percentage change.

Those two windows answer different questions. The 30-day total is what you report; the 7-day comparison is what tells you something changed recently, which is the more useful of the two for catching a runaway agent or a newly onboarded team.

Usage by group and by user​

Below the chart, a table breaks spend down, with tabs for groups and for users. Each row shows spend, the applicable budget, and utilization as a percentage.

Utilization is the column to sort on. A group at 101% has already had requests refused, since budget enforcement denies rather than warns. Filters for under 80%, 80 to 100%, and over 100% are there to make that triage quick.

A group with spend but no budget is worth attention too: it means traffic is being attributed but nothing is capping it.

What the numbers are​

Spend is computed from each provider's reported token usage, priced against the active pricing catalog. So a figure here is only as accurate as the rates you have published, and a model with no rate does not appear as free traffic, it appears as refused requests. See Budgets and pricing.

Group figures follow inherited membership, so a user in a subgroup contributes to the parent group's total.

Next steps​

  • Budgets to change what any group or user may spend.
  • Tool usage for the equivalent picture of tool calls rather than model calls.