Control spend

Two different controls, often confused. A budget bounds a run. The governor stops work that has stopped making progress — which is usually the spend you actually wanted interrupted.

Budget and governor are not the same thing#

A budget
Set per goal. It bounds how much a run may consume, and it is an arbitrary line — good work and bad work hit it identically.
The governor
Always on, per task. It watches real progress against spend and halts a run that has stopped producing any. It stops the case you would have wanted stopped.

Finding where the tokens went#

Spend is attributed per task and measured rather than apportioned, so the figure against a task is what that task actually caused, not a share of a total divided up afterwards.

The scorecard shows this alongside the outcome, which is the pairing that matters: cost on its own is not a problem, and cost against nothing shipped is.

You are billed on the capability tier

Routing may serve a role from a different model on different days. The rate card is published per tier, so which vendor served a call changes our margin and not your bill.

FeaturesToken intelligenceToken optimisation here never depends on a human noticing a chart. The system observes, diagnoses, fixes, validates and learns, so waste disappears on its own.FeaturesModel routingConnect an AI account and never make another engineering decision about models. Routing chooses per call, and two invariants stop it ever trading your outcome for a cheaper bill.GuidesRead the scorecardA recovery rate on its own can be gamed by any system willing to declare victory. These are the numbers to read together, and the ones to be suspicious of alone.

See the rate card

Published per capability tier, served by the same code that meters the charge.

See pricing →