Spotify Wrapped for Claude Code. Analyzes your ~/.claude/projects/ session data and generates a visual HTML report with usage stats, token costs, coding patterns, and personalized insights.

Quick Start

npx claude-entropy

Requires: Node.js 14+ and Python 3.8+ on PATH.

What You Get

A self-contained HTML report with 20+ slides covering:

Sessions & hours coded with Claude, daily streaks, peak coding days
Lines of code generated across all projects
Token usage and estimated cost (input, output, cache)
Project breakdown with per-project stats
Prompting style analysis: length, specificity, effectiveness
Error patterns: taxonomy of 14 error categories
Retry loops: wasted tokens from stuck patterns
Communication tone: niceness score, swear tracking
Self-scoring bias: how accurately Claude rates its own work
Tool usage: misuse detection (Bash vs Read, etc.)
Coding personality: archetype based on your usage patterns
Percentile ranking: how you compare to other Claude Code users

Options

npx claude-entropy --author "Your Name"         # Display name (default: git user.name)
npx claude-entropy --tz 1                        # UTC offset (default: auto-detect)
npx claude-entropy --money 600                   # Subscription cost for ROI slide
npx claude-entropy --money-detail "3 Max"        # Subscription description
npx claude-entropy --sanitize                    # Anonymize project names for sharing
npx claude-entropy --publish                     # Publish to wrapped.buildingopen.org
npx claude-entropy --help                        # Show all options

Output

Generates ./wrapped.html in your current directory and opens it in your browser.

More Reports

For deeper analysis (prompt coaching, personality profiles, character studies):

npx claude-entropy-lab prompt-coach
npx claude-entropy-lab user-profile
npx claude-entropy-lab soul
npx claude-entropy-lab portrait

See claude-entropy-lab for details.

How It Works

Reads Claude Code session files from ~/.claude/projects/ (JSONL format)
Runs 10 pattern analyzers in parallel (pure Python, no pip dependencies)
Computes aggregated stats, percentiles, and a personality archetype
Generates a single self-contained HTML file with animated slides

All processing happens locally. No data is sent anywhere unless you use --publish.

Requirements

Node.js 14+ (for npx)
Python 3.8+ (for analysis, uses only stdlib)
Claude Code session data in ~/.claude/projects/

Override the data directory with CLAUDE_PROJECTS_DIR env var. Supports multiple directories with : separator (; on Windows):

CLAUDE_PROJECTS_DIR="/path/to/mac-sessions:/path/to/server-sessions" npx claude-entropy

Privacy

Your session data never leaves your machine. The --sanitize flag strips identifying data: project names, prompt examples, swear quotes, uncensored swear words, and machine names. The --publish flag uploads only the final HTML report (not raw session data) to a public URL, and always auto-sanitizes.