While building claudebill we ran it over our own Claude Code history: 81 sessions, 7 billion tokens, about $7,600 in API-equivalent spend. Three things we did not expect:
- One session cost $481. It was a long homelab automation task that kept a huge context alive for hours. Nobody noticed at the time, because on a Max plan nothing shows a dollar figure.
- One project was 96% of all spend. Everything else combined was 4%.
- Prompt caching saved about $44,000. Without it the same work would have cost six times more.
Claude Code already logs every token it bills. claudebill reads those logs and shows cost per session, project, git branch, model, and day, plus what prompt caching saved you. Retroactive, offline, zero dependencies. npx claudebill.
