What is UsageRank?
UsageRank is a public usage leaderboard for Claude Code and Codex users who want to compare how heavily they use coding agents. Its CLI reads local Claude and Codex session logs, totals token usage by model and day, then pushes those aggregates to a public ranking with token and estimated spend views. Historical usage can be backfilled, while scheduled pushes keep rankings current.
Why UsageRank works
A public AI usage ranking becomes a bad trade if joining means exposing prompts or manually counting every session. UsageRank reads the activity already stored in local session logs and sends only dates, models, token counts, and hashed session IDs, so developers can compare real usage without publishing what they asked or what the agent wrote. It turns private local activity into public totals instead of publishing conversation content.
UsageRank features
- Claude Code and Codex leaderboard. Users are ranked by tokens burned across Claude Code and Codex, with separate filters for each tool and views for token usage or estimated spend.
- Automatic local log scanning. The CLI scans Claude project logs and Codex session logs, then calculates messages, sessions, token counts, and daily aggregates without asking users to enter usage manually.
- Historical backfill. A full push can send every day still present in local logs, letting new members populate previous activity instead of starting their ranking from zero.
- Scheduled updates. UsageRank can install a launch agent on macOS or a systemd user timer on Linux to push updated totals on a recurring schedule. Windows scheduling is not supported yet.
- Aggregate only uploads. Only dates, models, token counts, and hashed session IDs leave the machine. Prompts and model responses stay local.
- Model activity breakdowns. The activity view shows token usage over time and the share of total spend attributed to individual models.
Who UsageRank is for
- Developers using Claude Code heavily who want a public measure of how much agent work they are running across their local sessions.
- Codex users who want their token activity ranked against other users without uploading prompts or generated code.
- Developers switching between Claude Code and Codex who want one public view that combines usage from both local log sources.
- AI coding enthusiasts who enjoy competing on usage totals and want their ranking kept current automatically.
Similar micro SaaS ideas you can build
- Open source contributor league. A local Git tool that turns commit counts, active days, languages, and merged work into opt in public rankings. Repository contents stay private, so developers can compete on contribution activity without exposing proprietary code.
- Local AI lab leaderboard. A tracker for developers running models on their own hardware that publishes aggregate inference volume, model mix, and compute time from local logs. It gives home lab users a way to compare how much they run without uploading prompts or outputs.
- Security practice rankings. A desktop tracker for people doing CTFs and security labs that converts local practice sessions into public streaks, time spent, and challenge counts. Participants can compare consistency while keeping commands, notes, and solutions private.
