# TokenTrust

> The database of verified AI coding-agent token usage. Every number is labelled with where it came from.

## Trust tiers
- Provider-verified (tier A): read hourly from the provider's own usage API (Anthropic organization usage report) with an Admin key or a personal API key, used only to read usage. Ranked on /?board=api.
- Log-uploaded (tier C): per-response token counts from local Claude Code / Codex logs, uploaded by the tokentrust CLI with provider response ids and checked by fraud rules. Ranked on /. Checked, not proven.

## Pages
- /: leaderboard (last 30 days). Categories via ?board=: total (Claude Code + Codex, default), claude-code, codex, mom (month-over-month growth), api (API keys)
- /{handle}: developer profile; /{handle}.md for markdown

## API (public, no auth)
- GET /api/v1/devs: both leaderboards
- GET /api/v1/devs/{handle}: one profile with daily, monthly and per-model usage

Token counts are exact integers. Costs are API list-price equivalents in USD, not what the developer paid.