claude-history-explorer 0.2.1

CLI tool to explore Claude Code conversation history

Claude History Explorer

A Python CLI tool to explore, search and visualise your Claude Code conversation history. The history is stored locally at ~/.claude/projects/ and this tool turns raw JSONL files into searchable conversations and insights about your coding journey.

Highlights

• Story Generation - Analyzes sessions to create narratives about work patterns, personality traits, and collaboration style
• Concurrent Claude Detection - Identifies when you've used multiple Claude instances in parallel
• Rich Terminal UI - Tables, panels, sparklines, and syntax highlighting
• Multiple Export Formats - JSON, Markdown, plain text
• Regex Search - Search across conversations with Python regex plus a basic ReDoS safety guard
• Read-Only by Design - Never modifies your Claude history files (why trust this?)
• Streaming JSONL - Line-by-line parsing handles large files efficiently

Example Output

📖 Keyboardia Project Story
============================

📅 3 days of development
🤝 heavy delegation (16 agents, 2 main sessions)
⚡ 1873 messages at 25.0 msgs/hour
🎯 steady, productive flow with marathon sessions
🎭 agent-driven, deep-work focused, high-intensity
🔀 Used up to 3 Claude instances in parallel

💡 Key insight: Most productive session: 1530 messages

Installation

Using uv (recommended):

# Install directly
uv tool install .

# Or run without installing
uv run claude-history --help

Using pip:

pip install .

Usage

Tip: Every command supports --example to show usage examples: claude-history search --example

projects

List all Claude Code projects sorted by last use.

claude-history projects
claude-history projects -n 10        # Limit to 10 projects

Option	Description
-n, --limit	Maximum projects to show (default: 20)

sessions

List sessions for a project. Search is a partial match on project path.

claude-history sessions myproject
claude-history sessions "Documents/work" -n 5

Option	Description
-n, --limit	Maximum sessions to show (default: 20)
--head	Show most recent sessions (default)
-t, --tail	Show oldest sessions

show

Display messages from a session. Session ID can be a partial match.

claude-history show e5c477f0
claude-history show e5c477f0 -n 100
claude-history show e5c477f0 --raw
claude-history show e5c477f0 -p myproject

Option	Description
-p, --project	Limit search to specific project
-n, --limit	Maximum messages to show (default: 50)
--head	Show first messages (default)
-t, --tail	Show last messages
--raw	Output raw JSON

search

Search for a regex pattern across all conversations.

claude-history search "TODO"
claude-history search "error.*fix"
claude-history search "bug" -p myproject
claude-history search "API" -c
claude-history search "function" -C 200

Option	Description
-p, --project	Limit search to specific project
-c, --case-sensitive	Case-sensitive search
-n, --limit	Maximum results to show (default: 20)
-C, --context	Characters of context around match (default: 100)

export

Export a session to JSON, Markdown, or plain text.

claude-history export e5c477f0 -f markdown -o session.md
claude-history export e5c477f0 -f json -o session.json
claude-history export e5c477f0 -f text
claude-history export e5c477f0 -p myproject -f json

Option	Description
-p, --project	Limit search to specific project
-f, --format	Output format: json, markdown, text (default: markdown)
-o, --output	Output file (default: stdout)

info

Show Claude Code storage location and usage statistics.

claude-history info

stats

Show detailed statistics including message counts, duration, storage size, and agent usage.

claude-history stats
claude-history stats -p myproject
claude-history stats -f json

Option	Description
-p, --project	Show stats for specific project only
-f, --format	Output format: table, json (default: table)
--show-worktype	Include work type classification

summary

Generate a comprehensive summary with insights and charts.

claude-history summary
claude-history summary -p myproject
claude-history summary -f markdown -o report.md

Option	Description
-p, --project	Generate summary for specific project
-f, --format	Output format: text, markdown (default: text)
-o, --output	Output file (default: stdout)
--show-worktype	Include work type distribution

story

Tell the story of your development journey with personality insights and patterns.

claude-history story
claude-history story -p myproject
claude-history story -f brief
claude-history story -f timeline -o journey.md

Option	Description
-p, --project	Tell story for specific project
-f, --format	Story format: brief, detailed, timeline (default: detailed)
-o, --output	Output file (default: stdout)

Generates narrative insights including project lifecycle, collaboration style, work intensity, personality traits, and timeline visualization with sparklines.

wrapped

Generate a shareable Wrapped URL containing your year's aggregate stats. The CLI encodes data into the URL and makes no network calls; the website decodes the URL to render it and uses no-store responses.

claude-history wrapped
claude-history wrapped -y 2024
claude-history wrapped -n "Your Name"
claude-history wrapped --raw
claude-history wrapped --decode "https://..."

Option	Description
-y, --year	Year to generate wrapped for (default: current year)
-n, --name	Display name to show on wrapped cards
--raw	Output raw JSON instead of URL
--no-copy	Don't copy URL to clipboard
-d, --decode	Decode and display a Wrapped URL

How Claude Code Stores History

Claude Code stores conversation history in:

~/.claude/projects/
├── -Users-username-path-to-project1/
│   ├── session-id-1.jsonl
│   ├── session-id-2.jsonl
│   └── agent-xyz.jsonl
├── -Users-username-path-to-project2/
│   └── ...

• Project directories are named with an encoded filesystem path. Unix paths look like -Users-username-path-to-project, Windows drive paths like C--Users-username-path-to-project, and UNC paths like --server-share-project.
• Decoding is ambiguity-aware: when possible, the tool probes existing filesystem components to preserve dots, spaces, underscores, and hyphenated names.
• Each .jsonl file is a conversation session
• Files prefixed with agent- are sub-agent conversations
• Each line in a JSONL file is a JSON object representing a message or event

Development

# Clone and install in dev mode
git clone <repo>
cd claude-history-explorer
uv sync

# Install Wrapped website dependencies first so Python↔TypeScript bridge tests run
cd wrapped-website
npm ci
cd ..

# Run Python tests/lint
uv run --locked pytest -q
uv run --locked ruff check .

# Run privacy-preserving smoke test against your local Claude corpus
uv run --locked python scripts/smoketest_local_corpus.py

# Run Wrapped website tests/typecheck/lint/audit
cd wrapped-website
npm test
npm run typecheck
npm run lint
npm audit --omit=dev
npm audit
cd ..

# Run the CLI
uv run claude-history --help

Documentation

• Changelog
• Architecture
• Domain model
• JSON schemas
• Wrapped V3 spec
• Trust model
• Lessons learned
• Release process

License

MIT