BURNBOOKBURNBOOK
Burnbook · Vol. 01 · Token Evidence Platform

HOW MUCH
HAVE YOU
BURNED?

Install the CLI.Find out.Compare.
87.2B Processed2.4B Fresh · 84.8B cache readLive accepted evidence · refreshed within five minutes
Claim your handle →See the burn log ↓
// live demo — fabricated data, real component
Token Burn Record

you

@you · burning since MAR 2025 · claude code + codex preview
DEVICE
ATTESTEDCLI-SIGNED · REPLAY-SAFE
2.4Blifetime tokens
147day streak
#23global rank
// the burn log — last 39 weeks
// this week's tracklist
01refactor auth middleware184M tk
02burrow-cli release pipeline96M tk
03security scan sweep71M tk

This card is fabricated demo data. Yours is collector-derived evidence — claim a handle and burn something.

HOW IT
WORKS

  1. 01
    npm install -g burnbook

    Install the MIT-licensed CLI from the public npm package. No account is required yet.

  2. 02
    burn login

    Approve the device in your browser. Your keys never leave your machine.

  3. 03
    burn repair

    Install Claude Code hooks and the local scheduler. The scheduler collects Claude Code plus preview Codex and Gemini CLI evidence into a private spool; hooks themselves never use the network.

  4. 04
    burn sync

    Review and upload pending content-free evidence whenever you want your Burnbook profile updated.

Coverage today: Claude Code is certified. Codex and Gemini CLI are automatic previews. Cursor and Antigravity accept content-free preview imports. OpenClaw, Hermes, Lovable, and Bolt are not collected yet—Burnbook never estimates missing token evidence or labels credit balances as exact model tokens.

THE PROMISE

You'll only show real numbers if nothing sensitive rides along with them. So here's exactly what crosses the wire — no fine print, no asterisk.

WE COLLECT
  • Token counts, per session
  • Model IDs (claude-opus-5, gpt-5-codex, …)
  • Timestamps
WE NEVER COLLECT
  • Your prompts
  • Your code
  • File contents, paths, or diffs
ZERO PROMPT LEAKAGE.
EVER.