The problem
Most teams know their total AI bill but not what one match, report or feature actually costs, which is the number pricing depends on.
How it works
A system-wide HTTPS proxy captures every API call into SQLite, and a Next.js dashboard rolls spend up by business unit, provider and feature, flagging anomalies.
Highlights
- token-count: an Electron menubar app intercepts all Claude API calls system-wide via a local HTTPS proxy, no code changes required.
- token-sense: a dashboard that tracks cost per business unit, the metric that actually drives AI pricing decisions.
- Auto-refreshing dashboard with spend broken down by feature, provider and custom dimension.
- Flags anomalous API calls in a dedicated table with the specific issue highlighted.