tokensburned

AI CODING ACTIVITY FOR GITHUB

Your AI coding receipt.

Count tokens across coding harnesses, keep prompts and code on your machine, and publish one live profile card.

Install for your harness Build your card
Collected
Token totals and model metadata
Ignored
Prompts, replies, code, repo names
Static examplesample data
Static TokensBurned profile card with fictional sample data
MIT licensedServerless API15 minute aggregatesNo profile commits

SUPPORTED HARNESSES

Pick your coding harness.

Each integration uses the strongest official capability available. The panel tells you when collection is native, OTLP based, or CLI assisted.

NATIVE HOOK

Claude Code

Precise session totals

Install from the TokensBurned marketplace, connect GitHub, then let the official SessionEnd hook upload a short aggregate snapshot.

    /plugin marketplace add Parsifal1986/TokensBurned
/plugin install tokensburned@tokensburned
/reload-plugins
/tokensburned:connect

    Optional history import reads only Claude Code JSONL usage fields for the approved date range.

    PROFILE CARD BUILDER

    Make the damage fit.

    Explore fictional samples locally. Your username changes only the card label and generated link; the preview never calls the API.

    Card size
    Card theme
    Elements
    Previewsample
    Static TokensBurned preview with fictional sample data

    Static fictional preview. Enter a GitHub username to generate the link; no API request is made here.

    THE DATA BOUNDARY

    Aggregate locally. Publish deliberately.

    Leaves your machine
    Token counts, harness, provider, model, hashed session ID, 15 minute bucket.
    Stays on your machine
    Prompts, responses, source code, repository names, transcript paths, API keys.
    Runs automatically
    A short best-effort lifecycle upload. No cron job, daemon, proxy, or Git commit.

    Document the damage.

    Open source, privacy bounded, and mildly judgmental.

    Install TokensBurnedRead the source