Titan Memory

🚀 Titan Memory MCP Server

A tool that provides a neural memory system for LLMs, capable of learning and predicting sequences while maintaining state through memory vectors. The MCP (Model Context Protocol) server offers tools suitable for Claude 3.7 Sonnet and other LLMs to maintain context states during interactions.

🚀 Quick Start

I'm aware that there are currently issues, and I'll fix them! Ideally, this should run in full autopilot mode in Cursor (or other Claude desktops) without human intervention, creating a "brain" independent of the LLM version.

✨ Features

• Perfect for Cursor: Currently, Cursor automatically runs MCP in lightweight mode. You can take your hands off the wheel and let your LLM gain new memories.
• Neural Memory Architecture: A transformer-based memory system capable of learning and predicting sequences.
• Memory Management: Efficient tensor operations with automatic memory cleanup.
• MCP Integration: Fully compatible with Cursor and other MCP clients.
• Text Encoding: Converts text input into tensor representations.
• Memory Persistence: Saves and loads memory states between sessions.

📦 Installation

# Clone the repository
git clone https://github.com/yourusername/titan-memory.git
cd titan-memory

# Install dependencies
npm install

# Build the project
npm run build

# Start the server
npm start

📚 Documentation

The Titan Memory MCP server provides the following tools:

help

Get help on available tools.

Parameters:

• tool (optional): Name of a specific tool to get help for.
• category (optional): Filter tools by category.
• showExamples (optional): Whether to show examples.
• verbose (optional): Show detailed information.

register

Register a new user with the server.

Parameters:

• username (required): Username.
• email (required): Email address.
• password (required): Password.

login

Log in to the server using registered credentials.

Parameters:

• username (required): Username.
• password (required): Password.

logout

Log out the current user from the server.

No parameters.

💻 Usage Examples

Basic Usage

// Example code for use in Claude 3.7 Sonnet:
const { TitanMemoryServer } = require('titan-memory-server');

const server = new TitanMemoryServer();
await server.register({
  username: 'yourUsername',
  email: 'yourEmail@example.com', 
  password: 'yourPassword'
});

await server.login({
  username: 'yourUsername',
  password: 'yourPassword'
});

// Perform other operations...

await server.logout();

🔧 Technical Details

The Titan Memory MCP server includes a sophisticated memory management system to prevent memory leaks and ensure efficient tensor operations:

1. Automatic Cleanup: Regularly cleans up unused tensors.
2. Memory Encryption: Securely stores memory states.
3. Tensor Validation: Ensures tensors have the correct shape.
4. Error Recovery: Gracefully handles tensor errors.

The Titan Memory MCP server adopts a modular architecture:

• TitanMemoryServer: The main server class that registers tools and handles requests.
• TitanMemoryModel: Implementation of the neural memory model.
• VectorProcessor: Handles input and text encoding.
• MemoryManager: Manages tensor operations and memory cleanup.

🤝 Contribution

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

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.