ClaudeMeter vs ccusage
They are not rivals. They measure two different things. If you are a Claude Code power user you probably want both.
ClaudeMeter
Reads the server-truth plan quota from claude.ai. Answers the question: “When is Anthropic going to cut me off, and how much weekly budget do I have left?”
brew install --cask m13v/tap/claude-meter
ccusage
Reads local Claude Code JSONL files to estimate token spend per session, per day, per project. Answers the question: “How many tokens did that refactor just cost me?”
npx ccusage
| Feature | ClaudeMeter | ccusage |
|---|---|---|
| What it measures | Plan quota Anthropic enforces (5h + weekly + extra) | Tokens consumed by Claude Code sessions |
| Data source | claude.ai internal usage endpoint | Local ~/.claude/projects JSONL files |
| Surface | macOS menu bar popover + CLI | CLI only |
| Auth model | Reuses claude.ai browser session via extension | None (reads local files) |
| Refresh cadence | Once per minute | On-demand |
| Knows when Anthropic will cut you off | Yes | No (local tokens are an estimate) |
| Knows how many tokens Claude Code burned | No (by design) | Yes |
| Windows / Linux support | macOS only today | Cross-platform |
| License | MIT | MIT |
| Cost | Free | Free |
When to use each
Use ClaudeMeter when:you are hitting “message limit reached” banners, your Claude Code run got killed halfway through, or you want to know whether you can safely start a multi-hour agentic loop without getting rate-limited mid-run.
Use ccusage when: you want to know which of your Claude Code sessions ate the most tokens, whether a particular prompt style is more expensive, or how your spend trends over time by project.
Why the numbers do not match
People regularly report ccusage showing “5% used” while Anthropic says they are rate-limited. That is expected: ccusage estimates tokens from local files, and Anthropic enforces a plan quota server-side using cost and rate heuristics we do not have visibility into. ClaudeMeter exists to bridge that gap.