📌Overview

Purpose: The Knowledge Graph Memory Server is designed to provide persistent memory capabilities for Claude, enabling it to remember user information across conversations and learn from past errors through a lesson system.

Overview: This framework utilizes a local knowledge graph to maintain entities, relationships, and lessons, allowing for efficient storage and retrieval of user-related data. By keeping track of observations and error patterns, the server aims to improve user interactions and enhance Claude's reliability in providing assistance.

Key Features:

• Entity Management: Create, modify, and delete entities in the knowledge graph, allowing it to hold unique identifiers, types, and related observations about users.

• Lesson Tracking: Capture lessons related to errors and solutions, including detailed information on error patterns, metadata, and steps for verification, thereby enabling continuous learning and improvement.

• API Tools: A robust set of tools for creating entities, establishing relationships, adding and removing observations, managing lessons, and searching the graph, ensuring comprehensive functionality and ease of use.

---

Knowledge Graph Memory Server

A basic implementation of persistent memory using a local knowledge graph, allowing Claude to remember user information across chats and learn from past errors through a lesson system.

Core Concepts

Entities

• Unique Name: Identifier for the entity.
• Entity Type: Classification (e.g., "person", "organization").
• Observations: List of facts about the entity.

Example:

{
  "name": "John_Smith",
  "entityType": "person",
  "observations": ["Speaks fluent Spanish"]
}

Relations

Define connections between entities, stored in active voice.

Example:

{
  "from": "John_Smith",
  "to": "Anthropic",
  "relationType": "works_at"
}

Observations

Discrete pieces of information about an entity.

• Stored as strings.
• Attached to specific entities.
• Atomic (one fact per observation).

Example:

{
  "entityName": "John_Smith",
  "observations": [
    "Speaks fluent Spanish",
    "Graduated in 2019",
    "Prefers morning meetings"
  ]
}

Lessons

Special entities that capture knowledge about errors and solutions.

Components:

• Unique Name
• Error Pattern (type, message, context)
• Solution Steps and Verification
• Metadata (severity, timestamps, frequency)

Example:

{
  "name": "NPM_VERSION_MISMATCH_01",
  "entityType": "lesson",
  "observations": [
    "Error occurs when using incompatible package versions",
    "Affects Windows environments",
    "Resolution requires version pinning"
  ],
  "errorPattern": {
    "type": "dependency",
    "message": "Cannot find package @shadcn/ui",
    "context": "package installation"
  },
  "metadata": {
    "severity": "high",
    "frequency": 1,
    "successRate": 1.0
  }
}

API

Tools

• create_entities: Create multiple new entities.
• create_relations: Create new relations between entities.
• add_observations: Add new observations to existing entities.
• delete_entities: Remove entities and their relations.
• delete_observations: Remove specific observations from entities.
• delete_relations: Remove specific relations.
• read_graph: Read the entire knowledge graph.
• search_nodes: Search for nodes based on a query.
• open_nodes: Retrieve specific nodes by name.

Lesson Management Tools

• create_lesson: Create a new lesson from an error and its solution.
• find_similar_errors: Find similar errors and their solutions.
• update_lesson_success: Update success tracking for a lesson.
• get_lesson_recommendations: Get relevant lessons for the current context.

File Management

Handles two types of files:

• memory.json: Stores entities and relations.
• lesson.json: Stores lesson entities.

Files are split if they exceed 1000 lines for performance.

Cursor MCP Client Setup

Steps to Integrate

1. Clone the Repository:

   git clone [repository-url]
   cd [repository-name]
   

2. Install Dependencies:

   pnpm install
   

3. Build the Project:

   pnpm build
   

4. Configure the Server:

   node /path/to/the/dist/index.js
   

5. Activate in Cursor:
   • Use keyboard shortcut Ctrl+Shift+P.
   • Type "reload window" and select it.

Usage with Claude Desktop

Setup for Different Environments

Docker

{
  "mcpServers": {
    "memory": {
      "command": "docker",
      "args": ["run", "-i", "-v", "claude-memory:/app/dist", "--rm", "mcp/memory"]
    }
  }
}

NPX

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    }
  }
}

System Prompt

Customize the interaction prompt for better memory utilization.

Building with Docker

docker build -t mcp/memory -f src/memory/Dockerfile .

License

Licensed under the MIT License, allowing usage, modification, and distribution under specified terms.

Acknowledgments

Special thanks to the original repository owner for base code enhancements. Model Context Protocol