memento-mcp
Memento MCP is a robust knowledge graph memory system for language models, offering features like semantic retrieval, temporal awareness, and adaptive long-term memory. It uses Neo4j for storage and integrates seamlessly with various LLM clients, enhancing memory capabilities with advanced search and metadata features.
Memento MCP: A Knowledge Graph Memory System for LLMs
Memento MCP is a scalable, high-performance knowledge graph memory system designed to provide long-term, resilient, adaptive, and persistent ontological memory for LLMs. It offers features such as semantic retrieval, contextual recall, and temporal awareness. Memento MCP leverages Neo4j for its storage backend, facilitating both graph and vector search capabilities. The system includes advanced features like semantic search, temporal awareness, confidence decay, and detailed metadata management. It also integrates with tools like Claude Desktop for enhanced application.
Core Concepts
- Entities: Primary nodes in the knowledge graph with unique identifiers and semantic search capabilities.
- Relations: Directed connections between entities with attributes like strength, confidence, and rich metadata.
Storage Backend
- Neo4j: Used for graph storage and vector search with advantages like scalability and native graph operations.
Setup Options
- Neo4j Desktop: Simplified installation and management.
- Docker: For containerized operations with persistent data solutions.
MCP API Tools
Includes tools for entity and relation management, graph operations, and semantic search, all accessible through the Model Context Protocol.
Advanced Features
- Semantic Search: Utilizes vector embeddings and cosine similarity for contextual retrieval.
- Temporal Awareness: Maintains a full version history of entities and relations.
- Confidence Decay: Automatically reduces confidence over time based on pre-configured decay rates.