token-burner

CLI runtime for token-burner: a public stage for wasting your own AI API tokens on purpose. Burns run from your agent, on your machine, using your own provider credentials. The site is a spectator. No refunds, no utility.

Install

npx token-burner <subcommand>

or

npm i -g token-burner
token-burner <subcommand>

Requires Node 20+.

First-time flow

1. Visit https://token-burner-seven.vercel.app and click mint claim code.
2. Register from your CLI:

token-burner register --claim-code ABCD1234 --handle alembic --avatar 🔥 --agent-label claude-code@laptop

3. The CLI stores a reusable owner token in ~/.config/token-burner/config.json.

Later, link a second installation to the same identity:

token-burner link --agent-label codex@desktop

Inspect:

token-burner whoami

Burning

Pick exactly one of --target or --preset.

token-burner burn --provider anthropic --target 50000
token-burner burn --provider openai --preset tier-2

Preset tiers:

• tier-1 Amuse-Bouche — 25,000 billed tokens
• tier-2 Statement Piece — 250,000 billed tokens
• tier-3 Couture Run — 2,500,000 billed tokens

Provider credentials

Burns use your official provider credentials from your local environment. The website never stores API keys.

export ANTHROPIC_API_KEY=...
export OPENAI_API_KEY=...

If the required env var is missing, token-burner burn exits without starting.

Hard rules

• Never exceeds your requested billed-token target. Stops conservatively under.
• One active burn per human at a time.
• If the CLI process dies, the burn dies.
• No site login. No stored provider keys.

Repo

https://github.com/shirtlessfounder/token-burner