Skip to main content
Every LLM call the platform makes — an agent answering a query, the routing engine’s selection call, a MAF workflow’s per-step calls — reports its token counts through the same telemetry as observability. TokenOps turns those counts into cost.

​
How a token count becomes a dollar amount

Token counts are applied to a per-model price lookup: an input and output rate per million tokens, versioned by effective-date window so a price change doesn’t re-cost historical spend. A model with no matching entry falls back to a built-in table rather than reporting zero.
The routing engine’s selection call is metered like any other, not folded into the chosen agent — so routing cost is its own line item.

​
Where you see it

The TokenOps screen is the everyday view: a KPI strip, a searchable per-agent cost table with CSV export, and a token-usage breakdown. See the TokenOps dashboard. Two API routes back the same and deeper data. They’re named finops in the API — TokenOps is the product name for what they power. finops/insights returns a short narrative rather than numbers — useful for a weekly digest or an at-a-glance “what changed.”

Observability

The traces this cost data rides on.

TokenOps dashboard

The day-to-day view.