Claude Memory MCP

🚀 Claude Memory MCP Server - A Persistent Memory Server for Claude Desktop Application

This project implements an optimized memory technology based on extensive research, providing persistent memory functionality for large language models and integrating specifically with the Claude desktop application.

🚀 Quick Start

Installation

1. Clone the repository:

git clone [repository-url]
cd claude-memory-mcp-server

2. Install dependencies:

pip install -r requirements.txt

3. Start the server:

python mcp/server.py

Integration with Claude Desktop

Add the following content to the configuration file of Claude Desktop:

{
    "mcp_server": {
        "host": "localhost",
        "port": 5000,
        "memory_file_path": "path/to/memory.json"
    }
}

Running Tests

pytest

✨ Features

• Hierarchical Memory Architecture: Short-term, long-term, and archival memory levels.
• Multiple Memory Types: Supports dialogue, knowledge, entity, and reflection memories.
• Semantic Search: Retrieves memories based on semantic similarity.
• Memory Consolidation: Automatically merges short-term memories into long-term memories.
• Memory Management: Retains and forgets memories based on importance.
• Claude Integration: Can be directly used with the Claude desktop application.
• MCP Protocol Support: Compatible with the Model Context Protocol.

🔧 Technical Details

Architecture

The MCP server adopts a functional domain-based architecture, including the following components:

┌─────────────────────────────────────────────────────────┐
│                   Claude Desktop                        │
└───────────────────────────┬─────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────┐
│                     MCP Interface                       │
│  ┌─────────────────┐  ┌─────────────────┐  ┌──────────┐ │
│  │ Tool Definition  │  │ Request Handler │  │ Security │ │
│  └─────────────────┘  └─────────────────┘  └──────────┘ │
└───────────────────────────┬─────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────┐
│                Memory Domain Manager                    │
├─────────────────┬─────────────────┬────────────────────┤
│  Episodic Memory │  Semantic Domain │  Temporal Domain │
├─────────────────┴─────────────────┴─────────────────────┤
└                                                ┓

Memory File Structure

Memory data is stored in JSON format, as shown in the following example:

{
    "metadata": {
        "version": "1.0.0",
        "timestamp": "2023-10-26T12:00:00Z"
    },
    "memory_entries": [
        {
            "id": 1,
            "content": "The weather is nice today. Let's go for a walk in the park.",
            "context": ["weather", "park"],
            "timestamp": "2023-10-26T12:05:00Z"
        },
        {
            "id": 2,
            "content": "Remember to bring an umbrella. It might rain this afternoon.",
            "context": ["weather", "umbrella"],
            "timestamp": "2023-10-26T12:10:00Z"
        }
    ]
}

Usage Instructions

Running the Server

python mcp/server.py

Available Tools

• Query Memory: query_memory
• Update Memory: update_memory
• Delete Memory Entry: delete_memory_entry

Project Structure

claude-memory-mcp-server/
├── mcp/
│   ├── server.py          # MCP server implementation
│   ├── tools.py           # MCP tool definition
│   └── handler.py         # Request handler
├── security/
│   └── validation.py      # Input validation
└── utils/
    ├── embeddings.py      # Vector embedding tools
    └── schema.py          # Schema validation

Research Background

This implementation is based on extensively researched persistent memory technologies for large language models:

• Operating System Heuristic Memory Management: Similar to MemGPT's hierarchical memory architecture.
• Biologically Inspired Episodic Memory: Context-sensitive memory retrieval.
• Vector Embedding: Semantic search inspired by vector database methods.
• Self-Reflection: Achieves memory consolidation through regular reviews.

🤝 Contribution Guidelines

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is open-source under the MIT License. Please refer to the LICENSE file for details.

🙏 Acknowledgments

• Based on research on optimized memory technologies for large language models.
• Implemented the Model Context Protocol for integration with Claude.