claude-code-session-extraction

name: claude-code-session-extraction description: Extract and read previous Claude Code conversations from local storage. Use when reviewing past sessions, finding previous context, or debugging session history. user-invocable: false

Claude Code Session Extraction

Extract and read previous Claude Code conversations from local storage.

Session Storage Location

~/.claude/projects/<project-path-encoded>/<session-uuid>.jsonl

Path encoding rules:

• Drive letters: C: → C-
• Path separators: / or \ → -
• Example: /home/user/myproject → home-user-myproject
• Example: <LOCAL_PATH>/MyProject → <DRIVE>--<PATH>-MyProject

Platform locations:

Quick Commands

List all projects with sessions

ls ~/.claude/projects/

List sessions for a specific project (sorted by date)

ls -lt ~/.claude/projects/<PROJECT_PATH>/*.jsonl | head -20

Resume a session interactively

claude --resume

Resume a specific session by ID

claude --resume <session-uuid>

JSONL Format Reference

Each line is a JSON object with a type field:

Assistant message content blocks

{
  "type": "assistant",
  "message": {
    "content": [
      {"type": "thinking", "thinking": "..."},
      {"type": "text", "text": "..."},
      {"type": "tool_use", "name": "Bash", "input": {...}}
    ]
  }
}

Extraction Script

Basic extraction (markdown output)

import json
import sys
import os

def extract_session(session_file):
    """Extract a Claude Code session to readable format."""
    if hasattr(sys.stdout, 'reconfigure'):
        sys.stdout.reconfigure(encoding='utf-8')

    conversation = []
    summary_title = None

    with open(session_file, 'r', encoding='utf-8') as f:
        for line in f:
            try:
                obj = json.loads(line.strip())
                msg_type = obj.get('type')

                if msg_type == 'summary':
                    summary_title = obj.get('summary', 'Untitled')

                elif msg_type == 'user':
                    content = obj.get('message', {}).get('content', '')
                    ts = obj.get('timestamp', '')[:19] if obj.get('timestamp') else ''
                    # Skip system/command messages
                    if content and not content.startswith('<'):
                        conversation.append({
                            'role': 'USER',
                            'content': content,
                            'timestamp': ts
                        })

                elif msg_type == 'assistant':
                    content_blocks = obj.get('message', {}).get('content', [])
                    ts = obj.get('timestamp', '')[:19] if obj.get('timestamp') else ''
                    texts = []
                    tools = []

                    for block in content_blocks:
                        if block.get('type') == 'text':
                            texts.append(block.get('text', ''))
                        elif block.get('type') == 'tool_use':
                            tools.append(block.get('name', 'unknown'))

                    if texts:
                        full_text = '\n'.join(texts)
                        conversation.append({
                            'role': 'ASSISTANT',
                            'content': full_text,
                            'tools': tools,
                            'timestamp': ts
                        })
            except Exception:
                pass

    return summary_title, conversation


def to_markdown(summary_title, conversation):
    """Convert extracted conversation to markdown."""
    output = []
    output.append(f"# {summary_title or 'Untitled Session'}\n")

    for msg in conversation:
        role = msg['role']
        ts = msg.get('timestamp', '')
        tools = msg.get('tools', [])
        content = msg['content']

        output.append(f"## {role}")
        if ts:
            output.append(f"*{ts}*")
        if tools:
            output.append(f"**Tools used:** {', '.join(tools)}")
        output.append(f"\n{content}\n")
        output.append("---\n")

    return '\n'.join(output)


def get_projects_dir():
    """Get the Claude Code projects directory for the current platform."""
    home = os.path.expanduser('~')
    return os.path.join(home, '.claude', 'projects')


def list_sessions(project_path_encoded, limit=20):
    """List session files for a project, sorted by modification time."""
    projects_dir = get_projects_dir()
    project_dir = os.path.join(projects_dir, project_path_encoded)

    if not os.path.exists(project_dir):
        return []

    sessions = []
    for f in os.listdir(project_dir):
        if f.endswith('.jsonl'):
            full_path = os.path.join(project_dir, f)
            mtime = os.path.getmtime(full_path)
            sessions.append((full_path, mtime))

    sessions.sort(key=lambda x: x[1], reverse=True)
    return [s[0] for s in sessions[:limit]]


if __name__ == '__main__':
    if len(sys.argv) < 2:
        print("Usage: python extract_session.py <session-file.jsonl>")
        sys.exit(1)

    title, convo = extract_session(sys.argv[1])
    print(to_markdown(title, convo))

One-liner for quick preview

# Preview first few exchanges from a session
python3 -c "
import json,sys
for l in open(sys.argv[1],'r',encoding='utf-8'):
  o=json.loads(l)
  if o.get('type')=='user' and not o.get('message',{}).get('content','').startswith('<'):
    print('USER:',o['message']['content'][:200])
  elif o.get('type')=='assistant':
    for b in o.get('message',{}).get('content',[]):
      if b.get('type')=='text' and len(b.get('text',''))>20:
        print('CLAUDE:',b['text'][:200])
        break
" ~/.claude/projects/<PROJECT>/<SESSION>.jsonl | head -50

Common Workflows

Find a session by keyword

# Search session content for a keyword
grep -l "some keyword" ~/.claude/projects/<PROJECT>/*.jsonl

Get session summaries (titles)

# List all session titles for a project
for f in ~/.claude/projects/<PROJECT>/*.jsonl; do
  title=$(head -5 "$f" | grep '"type":"summary"' | python3 -c "import json,sys; print(json.loads(sys.stdin.read()).get('summary',''))" 2>/dev/null)
  if [ -n "$title" ]; then
    echo "$(basename $f): $title"
  fi
done

Export session to file

python3 extract_session.py ~/.claude/projects/<PROJECT>/<SESSION>.jsonl > conversation.md

Notes

• Thinking blocks are stored but contain signatures (encrypted) - readable content is in thinking field
• Tool results are in separate JSONL entries, linked via parentUuid
• Subagent sessions are in <session-uuid>/subagents/ subdirectories
• Global history (all prompts across projects) is in ~/.claude/history.jsonl
• Cross-platform: Use os.path.expanduser('~') in Python or $HOME/%USERPROFILE% in shell