Claude Memory

🚀 Claude Memory MCP Server

A Model Context Protocol (MCP) server implementation that provides persistent memory functionality for large language models, specifically designed for integration into the Claude desktop application.

🚀 Quick Start

Installation

pip install .

Claude Integration

Add the following content to claudia.json:

{
    "mcp_server": {
        "host": "localhost",
        "port": 1234,
        "api_key": "your_api_key"
    }
}

Run the Server

python -m mcp.server

Run Tests

pytest

✨ Features

• Hierarchical Storage 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: Memory retention and forgetting based on importance.
• Claude Integration: Ready for integration with the Claude desktop application.
• MCP Protocol Support: Compatible with the Model Context Protocol.

📚 Documentation

Overview

This project implements the optimal memory technology from extensive research areas. It provides a standardized way for Claude to maintain persistent memory across conversations and sessions.

Architecture

The MCP server follows a functional domain-based architecture, consisting of the following components:

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

Research Background

This implementation is based on a comprehensive study of current LLM persistent memory technologies:

• OS-inspired Memory Management: Similar to MemGPT's hierarchical storage architecture.
• Biologically-inspired Chained Memory: Context-sensitive memory retrieval.
• Vector Embeddings: Semantic search inspired by vector database methods.
• Self-reflection: Memory consolidation through regular reviews.

🤝 Contribution

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

📄 License

This project is under the MIT license. For details, please refer to the LICENSE file.

🙏 Acknowledgments

• Based on research on optimized memory technologies for LLMs.
• Implements the Model Context Protocol for integration with Claude.