Overview
DuckDB Knowledge Graph Memory MCP is an MCP server that implements knowledge graph-based persistent memory using DuckDB as the storage backend. This enables AI agents to store, query, and retrieve structured knowledge with graph relationships, providing enhanced context awareness and reasoning capabilities.
Features
Knowledge Graph Storage
- Entity Management: Store and manage entities with attributes
- Relationship Modeling: Define connections between entities
- Graph Queries: Navigate knowledge graphs efficiently
- Semantic Search: Find related information by meaning
Memory Operations
- Store Knowledge: Save facts, entities, and relationships
- Query Memory: Retrieve relevant information on demand
- Update Entities: Modify existing knowledge
- Delete Obsolete Data: Remove outdated information
DuckDB Backend
- Fast Queries: In-memory analytical performance
- Persistent Storage: Data survives between sessions
- SQL Interface: Powerful query capabilities
- Efficient Storage: Columnar compression
Use Cases
Conversational AI Memory
- Remember user preferences and context
- Track conversation history and topics
- Build user profiles over time
- Provide personalized responses
Knowledge Base Building
- Extract and store information from documents
- Build domain-specific knowledge graphs
- Connect related concepts and entities
- Enable semantic search over knowledge
RAG Enhancement
- Store structured facts alongside embeddings
- Combine vector search with graph traversal
- Improve retrieval accuracy
- Provide explainable answers
Agent Memory Systems
- Long-term memory for AI agents
- Remember tasks and outcomes
- Learn from interactions
- Build contextual understanding
Knowledge Graph Capabilities
Entity Types
- People, organizations, locations
- Documents, concepts, topics
- Events and timestamps
- Custom domain entities
Relationship Types
- Hierarchical (parent-child)
- Associative (related-to)
- Temporal (before-after)
- Causal (causes-effects)
Query Patterns
- Find entities by attributes
- Navigate relationships
- Discover connected entities
- Pattern matching
Integration
Works with AI assistants supporting MCP. Uses DuckDB for storage with automatic schema management. Compatible with vector databases for hybrid search approaches.
Pricing
Free and open-source. DuckDB is also free and open-source.
