baek-seunghyun/ai-coding-usage-card

AI Coding Usage Card is a self hosted GitHub profile card that shows off how many tokens, and roughly how much money, you have burned using AI coding tools like Claude Code, Codex, Gemini CLI, Copilot CLI, and OpenCode. It relies on ccusage, a tool that reads your local usage logs, and turns that data into an SVG image you commit into your own GitHub profile repository. There are no external accounts, no API keys, and no third party services involved. Everything is generated on your own machine from logs already sitting there. The script produces four card variants you can mix and match: a full card showing totals, cost in your chosen currencies, a breakdown of token types, activity stats, and per tool costs, a smaller half card with just totals and cost, a GitHub style contribution heatmap called the grass card showing 26 weeks of daily cost, and a combo card that merges the half and grass cards side by side. Costs are labeled as API equivalent, since subscription plans like Claude Max do not actually bill per token, so the number shown is what the same usage would have cost if paid for through the API instead. Setup requires Node 18 or newer and the GitHub CLI logged in with repository access. After forking the project and cloning it, you assign each computer a permanent unique device name, run a test command to generate the cards, then embed the resulting SVG files in your profile README. A daily scheduled task, using Task Scheduler on Windows or cron on macOS and Linux, reruns the script automatically so the card stays current. If you use more than one computer, each one gets its own device name and its own snapshot is merged into the combined card, so reruns do not double count usage. The author warns against copying AI tool log folders between computers, since usage cannot be reliably deduplicated that way. Colors, currencies, and layout are all adjustable by editing a short SVG template in the script. The README also documents a couple of gotchas: unescaped characters in dynamic text can break the SVG, and GitHub's image caching can make a freshly changed card appear stale for a few minutes. The project is released under the MIT license.