token-counter

polished prototype

Model Development

A stdlib-only Python CLI + local dashboard that parses Claude Code transcripts and estimates API-equivalent cost per model, day, and project. It includes cache-read/cache-write pricing, transcript deduping, and a plan-value comparison.

Pythonzero-depscost analytics

What it does

Models LLM cost from recorded usage: cache reads at 0.1×, cache writes at 1.25–2×, and regular input/output against a pinned pricing snapshot. It is an estimate, not a bill, and stays local.

Talk about work like this →