TokensBurned

TokensBurned tracks AI coding activity for developers who want usage statistics on their GitHub profile. Its client processes records locally, but the profile cards depend on a hosted service that receives aggregate data.

• Usage cards: Show token totals, a seven-day trend, an activity heatmap and tool breakdowns, with optional streak, cached-input and rank badges.
• Reported token counts: Track counts supplied by coding tools rather than estimates from prompt length or cost, with separate labels for the tool, provider and model.
• Local collection and history: Collect usage from Claude Code, Codex, Cline, OpenCode and Gemini CLI, including supported history imports; GitHub Copilot CLI supports live capture without history backfill.
• Publication controls: Cards stay private until you explicitly publish them, and display settings can't reveal hidden activity; a browser card builder previews designs with fictional data and supports light, dark and viewer-matched themes.

Usage uploads contain aggregate token and request counts, tool, provider and model labels, and activity dates and hours. They exclude prompts, responses, source code, repository names, file paths and credentials. An unrecognized provider endpoint may appear as a hostname in attribution.

The open-source client uses the MIT license and requires Node.js. Background startup is supported on macOS and Linux. Connecting and publishing require a GitHub account and an internet connection. Collection depends on the coding tool's version and available usage records; Cursor and Aider need an integration to supply observed usage rather than automatic capture.