angrysky56/kng-mcp-server

KNG MCP Server

Knowledge-Narrative-Graph MCP Server - A temporal narrative memory service for AI systems based on the Random Tree Model (RTM).

Overview

The KNG MCP Server implements a sophisticated memory organization system inspired by human memory hierarchies. It organizes information into temporal "books" and "shelves" that compress and preserve knowledge across different time scales (minute → hour → day → week → month → year).

Key Features

• Temporal Hierarchies: Information is organized across multiple time scales with intelligent compression
• Persistent Themes: Important concepts and narratives persist across time boundaries
• Smart Organization: Automatic categorization into active, recent, archived, and reference shelves
• Graph-Based Storage: Built on knowledge graphs with rich node and edge relationships
• MCP Integration: Exposed as Model Context Protocol tools for seamless AI integration

Architecture

The service is built on three core concepts:

1. Books: Temporal containers that hold related knowledge graphs organized by time
2. Shelves: Categories that organize books by their temporal relevance and purpose
3. Persistent Themes: Key concepts that carry forward through time boundaries

Available MCP Tools

• kng_create_book - Create a new temporal book from a saved graph
• kng_get_library_shelves - Get all library shelves organized by category
• kng_get_books_on_shelf - Get all books on a specific shelf
• kng_search_books - Search for books by query and tags
• kng_get_book - Get a specific book by ID
• kng_update_book - Update a book when its underlying graph changes
• kng_delete_book - Safely delete a book while preserving themes
• kng_get_library_stats - Get library statistics for monitoring

Installation

# Install dependencies
npm install

# Build the server
npm run build

# Run in development mode
npm run dev

# Start the server
npm start

Usage

As an MCP Server

The server runs as a stdio-based MCP server. Configure your MCP client to connect to it:

{
  "kng-mcp-server": {
    "command": "node",
    "args": ["path/to/kng-mcp-server/dist/stdio.js"],
    "transport": "stdio"
  }
}

Creating a Book

// Example graph structure
const graph = {
  metadata: {
    id: "graph-001",
    title: "AI Research Notes",
    description: "Research on temporal memory systems",
    tags: ["research", "memory", "AI"],
    created: new Date(),
    modified: new Date(),
    version: "1.0"
  },
  nodes: new Map([
    ["node1", { id: "node1", label: "RTM Model", type: "concept" }],
    ["node2", { id: "node2", label: "Temporal Compression", type: "theme" }]
  ]),
  edges: new Map([
    ["edge1", { id: "edge1", source: "node1", target: "node2", label: "implements" }]
  ])
};

// Create book with 'day' time scale
const result = await kng_create_book({
  graph: graph,
  timeScale: "day"
});

Configuration

The service uses RTM-inspired parameters:

• maxBranchingFactor: Maximum children per temporal node (default: 4)
• maxNarrativeDepth: Maximum temporal hierarchy depth (default: 6)
• compressionTargets: Target compression ratios per time scale
• continuityRules: Rules for determining what persists across time boundaries

Memory Organization

Shelf Categories

1. Active: Currently developing stories and themes (< 1 day old)
2. Recent: Recently completed narratives (1-7 days old)
3. Reference: Persistent themes that span long periods
4. Archived: Older narratives preserved for reference (> 30 days)

Temporal Scales

• Minute: No compression, full detail
• Hour: Slight compression (1.2x)
• Day: Moderate compression (2x)
• Week: Significant compression (4x)
• Month: High compression (8x)
• Year: Maximum compression (16x)

Development

# Type checking
npm run type-check

# Build
npm run build

# Development with hot reload
npm run dev

Integration with Aura Framework

This MCP server is designed to integrate with the Aura AI framework, providing:

• Enhanced search capabilities through structured narratives
• Deeper emotional context via persistent themes
• Enriched user profiles with long-term memory
• Automated memory organization with intelligent archiving

License

MIT

Contributing

Contributions are welcome! Please ensure all TypeScript types are properly defined and follow the existing code structure.