Every coding-agent CLI quietly writes a JSONL transcript of each request to disk, and almost nobody reads it. ccusage's bet is that the data needed to understand your spend is already on your machine—you just need something to add it up. It never calls an API, never phones home, and runs fully offline against logs Claude Code and a dozen other agents already produce.
What Sets It Apart
- Reads existing local JSONL logs instead of scraping a billing dashboard, so the numbers reconcile with what the agent actually sent—and keep working when you're offline.
- Aggregates the same data several ways: daily, weekly, monthly, per-session, and per 5-hour block matching Claude's billing windows, so you see both trends and the single window that blew up.
- Spans more than Claude Code—one unified report covers Codex, Gemini CLI, Copilot CLI, OpenCode and others, which matters once you juggle several agents.
- Custom pricing overrides and JSON export let you plug it into your own dashboards or correct for negotiated rates.
Who It's For
Great fit if you run coding agents heavily and want a fast, scriptable read on cost without trusting—or waiting on—a vendor dashboard; npx ccusage and you have numbers. Look elsewhere if you need team-wide rollups, seat management, or a hosted UI: this is a single-user, local-first CLI, and its figures are estimates derived from logged tokens and public pricing, not an authoritative invoice.
