Anthropic Export Extractor

A production-grade C tool for extracting and organizing conversations from Anthropic Claude JSON exports into human-readable markdown files with structured artifact management.

Author: Richard Tune rich@quantumencoding.io Company: QUANTUM ENCODING LTD License: MIT

Overview

The Anthropic Export Extractor transforms massive, monolithic conversations.json exports from Anthropic Claude into a clean, organized directory structure with:

• Markdown files for each conversation (human-readable format)
• Artifact extraction (embedded files, code snippets, attachments)
• JSON manifests for each conversation (machine-readable metadata)
• Timestamped output directories for versioning and audit trails

Features

• ✅ Production-Grade Architecture: Built with the library-first pattern (core logic in libjson_parser.a)
• ✅ RFC 8259 Compliant: Full JSON parsing with proper error handling
• ✅ Large File Support: Handles 300MB+ JSON exports efficiently
• ✅ Artifact Management: Extracts embedded content to organized artifacts/ directories
• ✅ Metadata Preservation: Captures UUIDs, timestamps, and conversation structure
• ✅ Cross-Platform: Builds on Linux, macOS, and Windows (with minor adjustments)

Build Instructions

Prerequisites

• GCC or Clang compiler
• GNU Make
• Standard C11 support

Building

make all          # Build all binaries
make rebuild      # Clean and rebuild from scratch
make help         # Display usage information

This produces:

• anthropic_export_extractor - The main extraction tool
• json_parser - JSON validation and pretty-print utility
• libjson_parser.a - The reusable JSON parser library

Usage

Getting Help

./anthropic_export_extractor --help    # Display full help menu
./anthropic_export_extractor -h        # Short version
make help                              # Via makefile

Basic Extraction

./anthropic_export_extractor conversations.json

Output Structure

The tool creates a timestamped directory with the following structure:

extracted_conversations_2025-09-30_17-08-37/
├── Conversation_Name_abc12345/
│   ├── Conversation_Name.md        # Human-readable markdown
│   ├── manifest.json               # Machine-readable metadata
│   └── artifacts/                  # Extracted files/attachments
│       ├── code_snippet.py
│       ├── diagram.svg
│       └── data.csv
├── Another_Conversation_def67890/
│   ├── Another_Conversation.md
│   ├── manifest.json
│   └── artifacts/
└── ...

Output Format

Each conversation directory contains:

1. Markdown File (*.md): Complete conversation history with:

   • Conversation metadata (UUID, timestamps)
   • Message-by-message transcripts
   • Sender identification (human/assistant)
   • Links to extracted artifacts

2. Manifest File (manifest.json): Structured metadata including:

   • Conversation metadata
   • List of artifacts and their types
   • Statistics (message count, artifact count)
   • External file references

3. Artifacts Directory: Extracted embedded content:

   • Code files
   • Attachments
   • Images
   • Data files

Example Output

═══════════════════════════════════════════════════════
   ANTHROPIC EXPORT EXTRACTOR
═══════════════════════════════════════════════════════

Input: conversations.json (335544320 bytes)

Parsing JSON...
Found 615 conversations

Created root output directory: extracted_conversations_2025-09-30_17-08-37/

Extracting conversations:
───────────────────────────────────────────────────────
  [23] Build_quantum_library (msg:23 art:5 ext:0)
  [18] Implement_neural_network (msg:18 art:12 ext:2)
  [45] Debugging_session (msg:45 art:8 ext:1)
  ...
───────────────────────────────────────────────────────

✓ Extraction complete: 615/615 conversations processed
✓ Output directory: extracted_conversations_2025-09-30_17-08-37/

Architecture

This project follows the Library-First Pattern:

1. Core Library (libjson_parser.a): Self-contained, reusable JSON parsing engine

   • Zero external dependencies
   • Clean public API via json_parser.h
   • Full memory management and error handling

2. Application Binary (anthropic_export_extractor): Thin wrapper that:

   • Handles command-line arguments
   • Manages file I/O and directory creation
   • Links against the core library

This architecture enables:

• Code reuse across multiple projects
• Easy testing and validation
• Clean separation of concerns
• Maintainable, production-grade code

Development

Using the JSON Parser Tool

The included json_parser utility can validate and format any JSON file:

./json_parser --help                  # Show help
./json_parser data.json               # Parse and display JSON
./json_parser --validate data.json    # Validate only
./json_parser --pretty data.json      # Pretty-print JSON
./json_parser --compact data.json     # Minify JSON

Cleaning Build Artifacts

make clean        # Remove all build artifacts

Code Structure

• json_parser.c/h - Core JSON parsing library
• json_extractor.c - Main extraction logic
• main.c - JSON parser test suite
• Makefile - Build system

Use Cases

• Knowledge Management: Archive and organize AI conversation history
• Documentation: Extract project documentation from development conversations
• Research: Analyze conversation patterns and artifact usage
• Backup: Create searchable, human-readable backups of Claude conversations
• Migration: Transform Anthropic exports into other formats

Technical Specifications

• Language: C11
• Build System: GNU Make
• Dependencies: Standard C library only
• Memory Management: Manual allocation with comprehensive cleanup
• Error Handling: POSIX errno + custom error reporting
• File Format: UTF-8 text (Markdown, JSON)

Performance

• Tested with 300MB+ JSON exports
• Handles 600+ conversations efficiently
• Minimal memory footprint (streaming parser architecture)
• Fast execution on commodity hardware

Contributing

This is a production tool from QUANTUM ENCODING LTD. For inquiries, contact:

Richard Tune Email: rich@quantumencoding.io Company: QUANTUM ENCODING LTD

License

MIT License - See LICENSE file for details

---

Built with precision. Built for production. Built by QUANTUM ENCODING LTD.