Usage

What the agent has actually consumed — runs, model tokens, and recorded cost.

Agent runs

Conversations started

Tokens

Recorded cost

Over time
Agent runs bucketed across the selected window.
By trigger
Which plays the work came from
By model
Which engine ran it
By outcome
How those runs ended

Metered from agent runs — triggers, scheduled work, and anything filed through them. Chat turns are not billed per run, so they are reported as a conversation count only.