Usage and context

How much each session has used, and how much room it has left to think in — counted from the agent’s own record of the conversation, never guessed. There are no prices, for a reason worth reading.

What each session has used, and how much room is left

Agents charge by the word — “tokens” are roughly words, and the count is what everything else here is measured in. For each session you get: how many words it sent and received; how much of what it sent it had already sent before, and so did not pay full price for again, which in a long session is most of it; and how many times it went to the model at all.

Beside that is the one number people actually want: how full the session’s memory of the conversation is. When it is getting full the app says so, and it counts how many times the agent has had to squeeze that memory down to carry on. For a whole project there is the same picture for today, across every session at once.

All of it is read from the file the agent already writes as it works, a few lines at a time as they arrive. Nothing is scraped off the screen and nothing is estimated from how much text went past.

Why there is no dollar figure

There used to be one, and it was removed. The API price was real arithmetic and it still misled: almost everybody running this is on a subscription, and telling someone on a flat monthly fee that a session cost four figures states a number that never left their account.

The subscription figure cannot be computed at all — no plan publishes a token allowance or a per-token value, and the multiplier in “Max 5×” is a multiple of an allowance, not of a price. One number was misleading and the other was invented, so the app counts tokens and says nothing about money.

One session, in detail

⌘⇧I opens one session in detail: a timeline of requests, where the tokens went, a breakdown by model, the largest requests, which tools were called and which of them kept failing, and how the context window filled up over time. There is a context meter with a bloat warning, and the alerts panel reads its thresholds from the same place so the two can never disagree.