Receipts

🧾 Receipts

Track and share what you actually did with Claude Code.

Receipts reads your local Claude Code logs and turns them into a dashboard: tokens burned by the hour, a usage streak, hours saved, and a feed of every session. Click any session to mint a receipt card — a thermal-receipt-styled image, sized for Slack and Teams, that shows what you did in plain English along with the prompt you used.

Why this exists

"How many tokens are you burning?" is becoming a real — and bad — question from managers. Receipts answers it with substance instead of a number: here is the work behind the tokens. And because the receipt card carries the prompt, a teammate who sees it can copy what you did. Showing off becomes teaching.

The headline metrics are tasks done, streak, hours saved, and how many things needed no code at all — because the point is to show non-coders that their daily work (writing, spreadsheets, research, planning) is Claude Code work too. Tokens are the fun garnish, not the score.

Run

python3.13 -m venv .venv
.venv/bin/pip install -r requirements.txt
.venv/bin/python app.py

Then open http://localhost:4830. In Claude Desk, just launch it — the .claude-app.json manifest handles install and start automatically.

What it reads

Only ~/.claude/projects/**/*.jsonl — the logs Claude Code already writes on your machine. Nothing is uploaded. A receipt leaves your machine only when you copy a card and paste it somewhere yourself.

How to share

1. Click any session in Your receipts, or hit Today's receipt.
2. Copy as image → paste straight into a Slack or Teams message.
3. The prompt is printed on the card on purpose — that is how a teammate reproduces what you did.

Notes on the numbers

• A session counts as running now if its log file was touched in the last 10 minutes.
• Time saved is a deliberately conservative estimate derived from tool activity. It is labelled an estimate everywhere on purpose — over-claiming here would poison the tool's credibility.
• Token totals include cache tokens (cache reads + cache creation), which are 90%+ of real usage. Tools that count only input_tokens + output_tokens undercount by roughly 100x once prompt caching is on.

Roadmap

• macOS menu-bar widget — live tokens/hour without opening the dashboard.
• Opt-in team view — patterns across a team, never a rank-ordered leaderboard.