Release Date: January 15, 2025

Claude Parser v2.0.0 is a complete rewrite that transforms the tool from a simple parser into a comprehensive disaster recovery and analysis platform for Claude Code conversations. This release introduces Git-like commands, a clean Python API, and a robust architecture built on LNCA principles.

• Full Git-like interface for navigating Claude conversations
• cg find - Search for files across all sessions
• cg checkout - Restore deleted files
• cg blame - Track file modifications
• cg reflog - View operation history
• cg show - Inspect specific operations
• cg reset - Time travel to any conversation state

• Composable hook runner for Claude Code integrations
• Pluggable executor architecture
• Support for all Claude hook events
• Environment variable configuration

• 30+ public functions across 15 domains
• Clean, intuitive imports
• Plain dict returns (no custom objects)
• Full Pydantic schema normalization

• Filtering: Message filtering by type, tool, content
• Watch: Real-time JSONL file monitoring
• Messages: Message processing utilities
• Models: Data models and utilities

• Direct SQL queries on JSONL files
• No intermediate database required
• Efficient handling of large sessions
• Schema inference and normalization

# v1 (old)
from claude_parser.main import ClaudeParser

# v2 (new)
from claude_parser import load_session, analyze_session

• Removed all god objects
• Functions return plain dicts, not custom classes
• Explicit imports required (no wildcard imports)
• New domain-based organization

• Correct path is ~/.claude/projects/ (not ~/.claude/code/conversations/)
• Project-specific files in /project/.claude/

• Every file <80 LOC for optimal LLM comprehension
• 100% framework delegation - no custom loops or error handling
• Single source of truth - one file per feature
• Pydantic everywhere - consistent schema handling

1. analytics - Session analysis
2. cli - Command line interfaces
3. discovery - File discovery
4. filtering - Message filtering
5. hooks - Hook system
6. loaders - Data loading
7. messages - Message utilities
8. models - Data models
9. navigation - Timeline navigation
10. operations - File operations
11. queries - SQL queries
12. session - Session management
13. storage - DuckDB engine
14. tokens - Token management
15. watch - Real-time monitoring

• 10x faster session loading with DuckDB
• 50% less memory usage with streaming
• Sub-second response times for most operations
• Efficient handling of GB-sized JSONL files

• Fixed schema mismatch in JSONL parsing
• Resolved UNION errors in SQL queries
• Corrected file path handling across platforms
• Fixed token counting for user vs assistant messages
• Resolved LOC violations in all modules

• typer - CLI framework
• rich - Terminal formatting
• duckdb - JSONL querying
• pydantic - Schema validation
• watchfiles - File monitoring (optional)

• Complete API reference
• User guide with real examples
• CLI command documentation
• Architecture overview
• Auto-deployed to GitHub Pages

1. Update imports to use new structure
2. Replace custom classes with dict operations
3. Use CG commands for recovery operations
4. Leverage new filtering capabilities

# v1
parser = ClaudeParser()
parser.load_session("file.jsonl")
parser.analyze_everything()

# v2
from claude_parser import load_session, analyze_session
session = load_session("file.jsonl")
analysis = analyze_session(session)

# Lost files? Recover them!
cg find "important.py"
cg checkout /path/to/important.py

from claude_parser import calculate_session_cost
cost = calculate_session_cost(input_tokens=100000, output_tokens=50000)

from claude_parser.filtering import filter_messages_by_type
user_messages = filter_messages_by_type(messages, "user")

• Claude Code community for feedback and testing
• Contributors who helped identify v1 limitations
• LNCA principles for guiding the architecture

• 500+ commits since v1
• 15 specialized domains
• 30+ public API functions
• 100% test coverage on critical paths
• 0 known critical bugs

• Advanced pattern matching in cg find
• Multi-session diff capabilities
• Enhanced hook integrations
• Performance optimizations for very large sessions
• Additional export formats

• v1 is now deprecated and will not receive updates
• v2 is the recommended version for all users
• Documentation available at https://alicoding.github.io/claude-parser/

Thank you for using Claude Parser! This v2 release represents months of work to create the ultimate disaster recovery tool for Claude Code users.