Neo4j GraphRAG

About

This MCP server enhances Neo4j to serve as a powerful backend for LLM-driven GraphRAG applications. It provides specialized tools for performing semantic vector searches, comprehensive fulltext searches using Lucene syntax, and innovative methods to combine search results directly within Cypher queries. Built on LiteLLM, it supports a wide range of embedding providers (OpenAI, Azure, Bedrock, Cohere, Ollama), allowing LLMs to intelligently interact with and retrieve information from a Neo4j graph database, making complex data exploration and retrieval more accessible for AI agents.

Key Features

• Semantic vector search using Neo4j vector indexes.
• Execute Cypher queries augmented by vector or fulltext search results.
• Keyword fulltext search with Lucene syntax on Neo4j fulltext indexes.
• 1 GitHub stars
• Multi-provider embedding support via LiteLLM for diverse AI models.
• Perform read-only Cypher queries for general graph traversal.

Use Cases

• Empowering LLMs to retrieve context-rich information from Neo4j for RAG applications.
• Integrating semantic and keyword search directly into AI agents' data querying workflows.
• Building conversational AI interfaces that can answer complex questions by combining graph traversal and advanced search.