Token spend: what makes it grow

You are billed a flat plan, but the AI subscription behind your agent has a usage budget. These three things consume it fastest.

1. Long chains of work

Each step in a long task re-reads everything that came before. A twenty-step job costs far more than twenty separate short questions.

2. Big material

Reading a long document, a large chat history or many web pages is the single most expensive thing an agent does. Point it at the specific file or page rather than at "everything in the folder".

3. Very long rules

Instructions you give the agent are re-read on every single step. A page of rules is paid for again and again. Keep them short and about behaviour — "always confirm before writing to a client" — rather than long lists of individual cases.

What does not help

Restarting the agent does not save anything — it costs, because the agent has to load its context again. Deleting history does not help either: the permanent memory is what makes the agent useful, and it is cheap to search.

Where do I see the numbers?

The dashboard shows spend per agent and per chat with a period selector. If one chat is consistently expensive, that is the one to change — usually by splitting a long-running task into separate topics.