Titan Memory

•
93
•

Empowers LLMs to maintain memory state across interactions using a neural memory system that learns and predicts sequences.

Titan Memory is a neural memory system designed for LLMs, enabling them to learn and predict sequences while maintaining state through a memory vector. Compatible with tools like Cursor, it provides a transformer-based memory architecture with efficient tensor operations and automatic memory cleanup. It is designed to allow LLMs to maintain a 'brain' independent of LLM version, by saving and loading memory states between sessions.

Key Features

01Text encoding for converting text inputs to tensor representations
02Compatible with MCP clients like Cursor
03Transformer-based neural memory architecture
04Memory persistence for saving and loading memory states
05Efficient tensor operations with automatic memory cleanup
0643 GitHub stars

Use Cases

01Creating AI agents that can learn and adapt over time
02Maintaining context across multiple interactions with an LLM
03Improving the consistency and coherence of LLM responses