Your AI token usage, as an API.

A collector reads every coding tool on every machine you use. One public API answers how many tokens, which models, which tools and which days.

npx github:bicrick/tokentracker init

Free and open source. Needs Node 22.13+ on macOS, Linux or Windows.

Usage over time.

Group by day, week, month or year, split by tool or model, filter by any range.

GET /api/v1/users/bicrick/usage?range=ytd&group=month&split=tool

Every model you've used.

Tokens, responses, share, and when you first and last used each one.

GET /api/v1/users/bicrick/models?range=ytd

    Across every tool.

    Claude Code, Codex, Cursor, OpenCode and more, added up across all your machines.

    GET /api/v1/users/bicrick/tools?range=ytd

    The numbers people ask about.

    Streaks, busiest day, today, this month, top tool and top model.

    GET /api/v1/users/bicrick/stats?range=ytd

    Build whatever you want.

    Plain JSON or CSV. No key, no SDK, cross-origin enabled. Every component on this page is one request.

    
          

    Read the API docs → llms.txt → OpenAPI →