Cost per run is the wrong number. Cost per resolved task, where it actually goes, and the four levers that move it without making the agent worse.
AI agent cost must be measured per resolved task, not per call; prompt caching is the biggest lever while input history dominates spend. Teams must optimize unit economics via model routing and fewer turns—token savings rank below scale profitability.
Book: AI That Ships
The series: AI in TypeScript — 5 books, from your first LLM call to agents in production — all five here
My project: Hermes IDE | GitHub — an IDE for developers who ship with Claude Code and other AI coding tools
Me: xgabriel.com | GitHub
The first cost question is always "what does a call cost". It is the wrong
AI agents do not become expensive all at once. They become expensive one small decision at a...
AI agent cost management belongs in runtime traces, evals, and harness policy, not monthly finance cleanup.
Cost is everything. In just about every agentic conversation, the three things that come up for...
The economics of autonomous agents depend less on the model and more on how much thinking, looping, and tool use you permit.
The setup Agentic coding tools are astonishingly capable, and astonishingly easy to run...
Build an AI coding agent cost ledger to track tokens, cache misses, idle waits, model routes, PR evidence, and expensive sessions…
