tokenstat
homepricingstatshelpsign in
tokenstat

Help center

Reading your usage

Your token totals are mostly cache reads, and a cache read costs a fraction of fresh input. These guides explain what a raw token count actually tells you, why a big context is cheaper than it looks, and how to read a cost figure that is a list rate rather than a bill.

3 guides in Reading your usage.

Why cache reads dominate your token count
Most of your tokens are cache reads, and they cost a fraction of fresh input. That is why a raw token number tells you almost nothing, and why a big context is not the expense people assume.
Which tokenstat command shows what
One archive, a dozen views over it. Which command answers which question, the four filters that work on all of them, and why every report can be piped as JSON.
Set a budget and put usage in your prompt
Two commands built for glancing rather than reading: a soft list-rate cap you can check at any time, and a single line you can hang off your shell prompt without slowing it down.