| name | cost-analysis |
| description | Analyze Claude Code session costs using ccusage. Shows token distribution, per-model breakdown, cache efficiency, and optimization recommendations. |
| argument-hint | ["session|daily|auto"] |
Cost Analysis
Analyze Claude Code token usage and costs for the SOLAR LINE project.
Quick Usage
Auto-detect (requires bun + ccusage):
cd ts && node --experimental-strip-types src/analyze-costs.ts --mode auto
Manual with ccusage pipe:
export PATH="$HOME/.bun/bin:$PATH"
bunx ccusage@17 session --offline --json --timezone Asia/Tokyo | \
node --experimental-strip-types ts/src/analyze-costs.ts --mode session
Daily summary:
export PATH="$HOME/.bun/bin:$PATH"
bunx ccusage@17 daily --offline --json --timezone Asia/Tokyo | \
node --experimental-strip-types ts/src/analyze-costs.ts --mode daily
What It Reports
- Token Distribution: Input, output, cache creation, cache read breakdown
- Per-Model Costs: Opus, Sonnet, Haiku token and cost breakdown
- Session Efficiency: Main vs subagent costs, expensive sessions, cache hit rates
- Recommendations: Actionable suggestions for reducing context waste
Key Metrics to Watch
| Metric | Good | Needs Attention |
|---|
| Cache hit rate | >95% | <90% |
| TodoWrite % of calls | <5% | >10% |
| Subagent avg tokens | <500K | >1M |
| Max session size | <2MB | >3MB |
Installation Note
ccusage v18+ requires the runtime: protocol (bun-native). Use v17:
bunx ccusage@17 [command] --offline --json
If bun is not installed:
curl -fsSL https://bun.sh/install | bash
export PATH="$HOME/.bun/bin:$PATH"
See Also
ideas/cost_efficiency_analysis.md — Full analysis write-up- CLAUDE.md "Context Efficiency" section — Operational guidelines
