← All tools
DEV-01Developer toolsReleased

tokenator

How hard are you actually pushing AI?

The tokenator landing page, headed “How hard are you actually pushing AI?”, with a live leaderboard below.

A collector runs on each machine where you use Claude Code, tails the transcripts it already writes, and ships per-message token counts to a central server. The dashboard rolls them up across the last hour, day, week, month, year and all time, estimates cost, and turns the total into a public rank, a tier and a shareable badge.

The collector is a single self-contained binary with no runtime dependencies. It dedupes by message and request ID, sends every thirty seconds, and sends an empty heartbeat when idle so the server can tell a quiet machine from a stuck one.

It also watches for format drift: if Anthropic ever changes the transcript shape, a new schema fingerprint appears in the dashboard as an early warning rather than as silently missing data.

What it does

  • Per-machine collectors, one static binary each
  • Rollups by hour, day, week, month, year and all time
  • Public leaderboard, tiers and an embeddable badge
  • Stale-machine detection via idle heartbeats
  • Transcript format-drift early warning
  • Your prompts never leave your machine, only counts

A closer look

The tokenator leaderboard listing developers by rank, tier, token count, percentile and streak.