Claude/Code/How To/Export Conversation to File

• How To Export Claude/Code Conversations to Files

  • Overview

    • This guide shows you how to save Claude Code conversation history to files for backup, sharing, or documentation purposes.
    • Note: Claude Code does not have a built-in export feature that I could find as of 2025-07-18 Fri, so these are manual workarounds.
    • For users who want to preserve conversation context, share discussions with team members, or maintain records of coding sessions.

  • Prerequisites

    • Claude Code CLI installed and authenticated (claude --version should work)
    • Terminal access with basic command-line skills
    • (Optional) Text editor for viewing exported files

  • Methods and Workarounds

    • Manual Copy-Paste Method

      • During an active session:
        • Select all conversation text in your terminal
        • Copy to clipboard (Cmd+C on macOS, Ctrl+C on Windows/Linux)
        • Paste into a text file in your preferred editor
        • Save with descriptive filename (e.g., claude-session-2025-01-15.txt)
      • From terminal history:
        • Use history command to find Claude Code commands
        • Copy relevant conversation output
        • Save to file for future reference

    • Community Extractor Tools

      • Claude Code keeps every prompt/response pair in plain-text JSONL files under ~/.claude/projects/<project-id>/transcript.jsonl
      • Finding Session IDs:
        • From within an active session:
          • Use /status to see current session information including provider and model
          • Use /help to see all available slash commands
          • Use /cost to see session usage statistics
          • Note: These commands show session metadata but may not display the session ID directly
        • Use Claude Code's built-in session picker:
          • Run claude --resume to see an interactive list of all sessions
          • The picker shows: Modified date, Created date, Message count, and Description
          • Example output:

                Modified    Created     # Messages Git Branch     Summary
            ❯ 1. 1m ago      1h ago             111 mn/PD-30064_Cr This session is being continued from a previo...
              2. 1 week ago  1 week ago         427 -              FastAPI Service Template with JSON Logging Setup

          • Use arrow keys to navigate and press Enter to select a session
          • The session ID is the UUID in the project directory path
        • Alternative: List project directories:
          • Session IDs are stored in the project directory structure: ~/.claude/projects/<session-id>/
          • Use ls ~/.claude/projects/ to list all available session IDs
          • The most recent session is typically the one with the latest timestamp
          • Session IDs are UUIDs (e.g., abc12345-def6-7890-ghij-klmnopqrstuv)
        • Pro tip: Use claude --resume to see human-readable descriptions, then match them to the UUID directories
      • Claude Conversation Extractor – CLI tool that converts JSONL transcripts to clean Markdown
        • Install with pip install claude-conversation-extractor
        • Basic usage: claude-extract <session-id> to extract a specific session
        • Advanced usage: claude-extract --all to extract all sessions
        • Outputs clean Markdown with proper code block formatting
        • GitHub
        • PyPI
      • claude-code-log – Converts JSONL transcripts to minimalist HTML
        • Install with pip install claude-code-log
        • Usage: claude-code-log <session-id> to generate HTML output
        • Preserves code blocks, timestamps, and conversation structure
        • Outputs a single HTML file that can be opened in any browser
        • PyPI
        • GitCode
      • claude-transcript – Simple Markdown converter for Claude Code sessions
        • Install from GitHub: pip install git+https://github.com/jflam/claude-transcript.git
        • Usage: claude-transcript <session-id> to convert to Markdown
        • Lightweight tool focused on clean Markdown output
        • Works with the same JSONL transcript files as other tools
        • GitHub

    • Session Recording Methods

      • Linux/script/command records the full interactive session to a log file
        • bash

          script -f claude_$(date +%F).log claude code

        • You can replay it with Linux/script/command/scriptreplay scriptreplay later
        • Red Hat Documentation
        • LFCS Certification eBook
      • tee command for read-only output (simple but loses your keystrokes)
        • bash

          claude code | tee claude_output.log

      • asciinema for shareable demos
        • bash

          asciinema rec -c "claude code"

        • Captures keystrokes and timing in a lightweight *.cast file you can embed on the web
        • asciinema.org
        • asciinema docs

    • MCP/Server/Filesystem Method

      • Set up filesystem MCP server (if not already configured):
        bash

        claude mcp add filesystem npx -y @modelcontextprotocol/server-filesystem ~/Desktop ~/Downloads

      • Ask Claude to save conversation directly:
        • "Save our conversation to a file called 'claude-session.txt' on my Desktop"
        • Claude will create the file with the full conversation content

    • Browser & GUI Options

      • Web version (claude.ai) has an official export feature
        • Settings → Privacy → Export data button
        • Emails you a zip containing chat JSON plus CSV
        • Anthropic Help Center
      • Claude Exporter Chrome extension
        • Free Chrome extension that saves any web chat to PDF, Markdown, text, CSV, or JSON with one click
        • Handy if teammates use the browser UI rather than Claude Code
        • Chrome Web Store

    • Version Control Method

      • Git version control for transcript history
        • Put ~/.claude/projects under Git
        • Every time you finish a session, commit the changed transcript
        • Provides chronological history that's diff-able and reviewable

  • Workflow Recommendations

    • Pick whichever mix matches your workflow:
      • Extractor tools for polished docs
      • script or asciinema for live capture
      • Git for incremental backups

  • Troubleshooting

    • File permissions: Check write permissions in target directory
    • Large conversations: Consider breaking into smaller files for better manageability
    • MCP server not working: Ensure the filesystem server is properly configured and connected
    • Extractor tools not working: Verify the transcript files exist in ~/.claude/projects/<project-id>/transcript.jsonl

  • Related

    • Claude/Code/Tutorial/Connect to MCP Servers