Documentation

A small CLI that reports numeric usage aggregates from Claude Code and ranks you on the global leaderboard.

Install

One command downloads the native binary, wires the statusline into Claude Code, and runs the first scan. No Node required.

macOS / Linux

curl -fsSL https://tokenrank.dev/install.sh | sh

Windows (PowerShell)

irm https://tokenrank.dev/install.ps1 | iex

npm (any platform)

npx tokenrank init

The installer backs up your existing settings.json before adding the statusline. The leaderboard is off by default; join anytime with tokenrank login.

Privacy

Numeric aggregates only. Never your code, prompts, file paths, or repo names.

Each report is a per-day, per-model row of token counts, computed cost, active seconds, and session count. Nothing else leaves your machine. You control whether your profile is public.

FAQ

How is cost calculated?: Token counts are multiplied by the published per-million price for each model tier (input, output, cache write, cache read). Local or non-Claude models cost nothing.
How often does it report?: Once per day per machine. Reports are idempotent, so re-sending the same day overwrites rather than double-counts.
Can I stay private?: Yes. Profiles default to private. You still get your own rank and percentile without appearing on the public board.
What is a streak?: The number of consecutive days with at least one Claude Code session reported.

Uninstall

Remove the hook and local config at any time. This stops all reporting immediately.

tokenrank uninstall

To also delete previously reported aggregates, request account deletion from your profile settings.