Infrastructure
Token Usage
How many tokens your API calls consume, which determines your AI bill.
Also written as token spend, token maxing
- Messages
- 103
- Members using it
- 49
- First seen
- Nov 22, 2025
What it means
The measure of text processed by an LLM, both input and output, billed per million tokens. Optimizing token usage means getting the same work done with fewer tokens through better prompts, smaller contexts, or cheaper models. Uncontrolled usage on large projects can cost hundreds of dollars daily.
How it's used here
A constant concern for members running agents at scale. Comes up when sharing token-saving skills, comparing model efficiency, complaining about reasoning token costs, or troubleshooting budget overruns. Members share techniques like the 'caveman' skill to cut usage.
Seen in the wild
“The only people who can pay for per-token usage are those who work for companies with ridiculous budgets and do not care about token maxing.”
“This skill has been saving my token usage, hope it helps someone: https://github.com/JuliusBrussee/caveman/blob/main/skills/caveman/SKILL.md”
“Pricey. But includes the token usage and is a ready-to-go harness, branded and aimed at people who just want something that gives them answers”