HN Showcase
Back to browse
GitHub Repository

See what Claude Code is really doing to your context — per-session & per-skill token/cost/time analysis and a Textual TUI over local transcripts. Find the prompt or skill quietly slowing you down. Python.

0 starsPython

Claude Code Session Trace/Browse Tool (Python)

by TheMadHatter76·Jun 25, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainNiche Gem

Reads existing transcripts with zero instrumentation — the bloat ratio metric exposes hidden token waste.

Strengths
  • •Wire-protocol parsing of existing JSONL files means no wrapper or capture step needed.
  • •Bloat ratio metric reveals how much standing context replays each turn.
  • •Textual TUI lets you drill into specific sessions and turns interactively.
Weaknesses
  • •Only works with Claude Code, not other AI coding assistants.
  • •Narrow audience limits broader adoption beyond Claude Code users.
Target Audience

Claude Code users, AI agent developers

Similar To

LangSmith · Helicone · Braintrust

Post Description

I have been testing different harnesses, agent memory tools, etc. and wanted to follow along with my claude sessions and analyze what files, tools, skills were burning tokens. Built this small tool to analyze claude sessions on disk.

The idea is basically a tui that allows you to select a project or session from claude, and then drill into it. Looking for token consumption, errors, or slowdowns. An easy way to get down to the nitty gritty of what happened at a specific time/turn.

I found it useful, so I thought I would share.

Similar Projects