The CLI (`npx @jaobrown/clog`) parses local Claude Code session data from `~/.claude/projects/` and computes stats: sessions, durations, token usage, project breakdowns, and streaks.

Running clog sync (optionally as a cron job in the background) pushes aggregate stats to a GitHub repo, and the web app at clog.sh aggregates everyone into a weekly leaderboard with individual profile pages.

Design decisions:

- All processing is local. Only aggregate metrics are published to github + leaderboard.

- GitHub is the sync layer — your stats live in a public repo you control, and repos with the tag are scooped up hourly.

- Sensitive project names can be redacted via clog redact while keeping totals intact.

- Supports both the modern sessions-index.json format and legacy JSONL parsing—immediately get 30+ days of usage stats up.

- Subagent activity is included in totals.

The leaderboard is still early — I'm looking for Claude Code users who want to try it.

Repo: https://github.com/jaobrown/clog Setup: `npx @jaobrown/clog init` Example profile: https://clog.sh/u/jaobrown