CHANGELOG.md

Changelog

All notable changes to megg will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.1.0] - 2026-01-17

Added

• Session State - New state() tool for ephemeral cross-session handoff
  • /megg-state skill for Claude Code
  • Auto-expires after 48 hours
  • Hard limit of 2k tokens to prevent bloat
• One-Command Setup - megg setup configures everything:
  • Registers MCP server automatically
  • Installs /megg-state skill
  • Configures SessionStart hook
  • Supports --uninstall to remove configuration
  • Supports --link for development symlinks
• Setup-Free Usage - Works immediately after npm install -g megg
• Improved CLI - Better help text and examples

Changed

• Simplified architecture from 9 tools to 6 core tools:
  • context(path?, topic?) - Load context chain + knowledge + state
  • learn(title, type, topics, content) - Add knowledge entry
  • init() - Initialize megg in directory
  • maintain() - Analyze and clean up knowledge
  • state(content?, status?) - Session state management
  • setup() - First-time machine configuration
• Skills now use folder structure: .claude/skills/megg-state/SKILL.md
• SessionStart hook includes state when active

Fixed

• Hook output format now properly includes additionalContext
• Token estimation more accurate for size thresholds

[1.0.0] - 2026-01-14

Added

• Initial release of megg v1 (v2 architecture)
• Core tools: context, learn, init, maintain
• Domain hierarchy system (not code folder based)
• Size-aware knowledge loading:
  • Full load under 8k tokens
  • Summary mode 8k-16k tokens
  • Blocked above 16k tokens
• Topic-based filtering for knowledge
• MCP server for Claude Code, Claude Desktop, and other MCP clients
• CLI for terminal usage

Philosophy

• Domain hierarchy for bounded contexts, not code folders
• Single knowledge.md per domain with type + topics
• Collaborative learning - agent asks before capturing
• Size thresholds prevent context bloat