lohani-mohit/tokenwise
5 stars · Last commit 2026-07-01
🪙 A local, zero-dependency token-cost profiler for Claude Code sessions. Finds what actually burned your tokens (big results that re-bill every turn, redundant reads, runaway subagents) and how to cut it.
README preview
<div align="center"> # 🪙 tokenwise **See exactly what a Claude Code session cost you — in tokens _and dollars_ — and how to spend less.** `tokenwise` runs a tiny local analyzer over your session transcript (main **and** subagents) and shows you the real cost drivers: big tool results that re-bill on every later turn, redundant file reads, runaway subagents, and cache efficiency — then turns that into concrete, prioritized cuts. The best part: the heavy parsing happens in a **zero-dependency Node script**, so Claude only reads a compact summary. The tool that tells you to stop dumping data into context doesn't dump data into context. [Install](#install) · [What you get](#what-you-get) · [How it works](#how-it-works) · [Why it's different](#why-its-different) · [FAQ](#faq)        </div> ---