CostClaw
Reads Claude Code logs locally, computes token spend, cache hit rates, and scores setup across six pillars, then ranks fixes by recoverable dollars.
CostClaw is a local audit tool that analyzes Claude Code session logs to identify token spend waste and provide actionable recommendations. It solves the problem of hidden cost inefficiencies in AI coding workflows by parsing logs on the user's machine and generating a report with recoverable spend estimates. The tool targets developers and engineering teams using Claude Code, particularly those in small to mid-sized companies or individual developers. It is positioned as a privacy-focused, on-premise solution that never uploads prompts or code, offering a free CLI and an optional hosted dashboard for Pro users.
Key features
- Local audit of Claude Code logs
- Identifies recoverable spend
- Scores setup across six pillars
- Ranks fixes by recoverable dollars
- Tracks cache hit rate per turn
- Flags sessions with high tool loops
- Detects model misrouting
- Identifies sessions that outgrow cache
- Provides actionable recommendations
- Privacy-focused: no uploads
- No social media activity detected
- No GTM activity detected
- Software developers
- Engineering teams
- Startups