Settings

Usage

What you have actually spent, by tokens, cache and session.

What it shows

A read-only dashboard over a date range: Total Tokens, Cache Hit Rate and Sessions.

Cache hit rate is the one worth watching. A long chat that keeps its context warm costs a fraction of one that re-sends it, so a low rate on heavy work usually means something is resetting the conversation rather than continuing it.

This is the meter. What you are allowed to spend, and what happens when it runs out, is on Plan.

Breakdowns

Below the cards, the same range is broken down three ways:

• By day, as a chart you can switch between metrics, to spot the day a long agent run happened.
• By model, so you can see what the expensive models actually cost you compared with the cheap ones.
• By project, expandable down to individual sessions, so a single runaway chat is findable.

The range you choose is remembered while the app is open, so leaving the page and coming back does not reset it.