claude-usage-analyzer

To analyze Claude Code token usage from local transcripts, run the build_usage_db.py script to generate a SQLite database from your ~/.claude data, then run generate_dashboard.py to view an interactive HTML dashboard showing token breakdowns by day, project, and model.

Yes, Claude Code session patterns and cache efficiency can be visualized without external dependencies by running a local analytics pipeline that ingests ~/.claude data, normalizes it into SQLite, and outputs a configurable HTML dashboard with zero external requirements.

A Claude Code usage analytics dashboard reveals token distribution across models, project activity levels, tool usage patterns, session insights, and cache efficiency metrics. It helps identify high-cost sessions and compare time ranges to optimize your workflows.

To track Claude Code project activity and tool patterns across date ranges, execute build_usage_db.py with your target dates to populate the SQLite database, then generate_dashboard.py to surface session patterns, token distribution, and workflow insights in an interactive view.

No, the Claude usage analyzer requires zero external dependencies and no API connections. It operates entirely locally by extracting data from ~/.claude transcripts, processing it with SQLite, and rendering a self-contained HTML dashboard.

The best way to identify high-cost Claude Code sessions is to generate a local analytics dashboard that breaks down token usage by day, project, and model, allowing you to compare time ranges and isolate sessions with inefficient cache usage or excessive token consumption.