jakeefr

jakeefr / prism

Public

Session intelligence for Claude Code - find extra token usage, why your sessions fail, and how to fix it.

anthropic claude claude-code cli developer-tools
19
1
100% credibility
Found Apr 14, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

PRISM is a local analysis tool that examines session records from an AI coding assistant to score efficiency, detect issues like wasted token re-reads and ignored rules, and suggest concrete fixes.

How It Works

1
🔍 Hear about PRISM

You notice your AI coding helper is using way too many tokens on repeat instructions and wonder why, then find PRISM which promises to reveal the hidden waste.

2
📥 Set it up easily

With one simple action, you add PRISM to your computer so it's ready to check your past sessions.

3
🔎 Scan your projects

You tell PRISM to look at the records from your recent AI coding sessions, and it quietly reads them all.

4
📊 Discover the problems

A colorful dashboard pops up showing letter grades for token use, rule following, and more, spotlighting exactly what's draining your budget.

5
💡 Get fix suggestions

PRISM's advisor points out specific lines to cut or rules to add in your instruction file, with clear reasons why.

6
✏️ Make the changes

You follow the simple tweaks, like trimming extra notes or moving rules to the right spot, and confirm they're applied.

✅ Enjoy smarter sessions

Your AI helper now follows rules better, uses fewer tokens, and gets work done faster without the hidden waste.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is prism?

Prism is a Python CLI tool that analyzes your Claude Code session files from ~/.claude/projects/, revealing why sessions burn tokens, ignore CLAUDE.md rules, or fail with tool loops. Run `prism analyze` for health scores across token efficiency, tool health, context hygiene, rule adherence, and continuity; `prism advise` delivers concrete CLAUDE.md fixes. It offers a TUI dashboard, HTML export, live session watching, and even Claude Code plugin integration—all local, zero telemetry.

Why is it gaining traction?

Unlike generic token counters, Prism surfaces root causes like CLAUDE.md re-read waste (e.g., 480% of session tokens) or retry loops, with diff-ready advice. The interactive TUI and browser dashboard make debugging sessions visual and fast, while live watch mode tracks active Claude Code runs in real-time. Devs love the "Mr. Tinkleberry test" automation for rule drift, turning opaque AI costs into actionable insights.

Who should use this?

Claude Code power users hitting token limits mid-project, backend devs debugging Bash tool failures or migration edits, frontend teams enforcing CLAUDE.md style rules that Claude ignores. Ideal for anyone running long artificial intelligence sessions where continuity breaks or context compacts unexpectedly.

Verdict

Try Prism if you're deep into Claude Code—its diagnostics are sharp and local, with solid docs and tests despite 12 stars and 1.0% credibility signaling early alpha status. Install via pip for quick wins, but watch for stability as it matures.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.