🔍 MCP Server - Vector Search

What is 🔍 MCP Server - Vector Search?

🔍 MCP Server - Vector Search is a Model Context Protocol server built with FastMCP that combines Neo4j’s graph database with vector search using embeddings. It enables intelligent semantic search across a knowledge graph through natural language queries, converting them into 1536-dimensional vectors via OpenAI (or a sentence-transformers fallback) and returning ranked results from a Neo4j vector index.

How to use 🔍 MCP Server - Vector Search?

Install with uv, clone the repository, set up a .env file with Neo4j credentials and an optional OpenAI API key, create a vector index in Neo4j, then run python main.py. The server exposes a single tool vector_search_neo4j(prompt) that accepts natural language queries and returns semantically similar nodes with similarity scores.

Key features of 🔍 MCP Server - Vector Search

• Combines Neo4j graph database with vector search
• Uses OpenAI embeddings or sentence-transformers as fallback
• Single powerful tool for semantic search
• Asynchronous, non-blocking I/O for high throughput
• Easy integration with Claude Desktop via MCP
• Dependency management with uv (10‑100x faster resolution)

Use cases of 🔍 MCP Server - Vector Search

• Semantic document retrieval in knowledge-graph-powered applications
• Natural‑language querying of structured and unstructured data stored in Neo4j
• Connecting AI assistants (e.g., Claude) to graph knowledge for context‑aware answers
• Intelligent search across enterprise knowledge bases that use graph and vector indexes

FAQ from 🔍 MCP Server - Vector Search

What does the vector_search_neo4j tool do?

It converts a natural‑language prompt into a 1536‑dimensional vector (using OpenAI or sentence‑transformers), searches the Neo4j vector index for the most semantically similar nodes, and returns ranked results with similarity scores.

How do I set up the Neo4j vector index?

Run the Cypher command: CREATE VECTOR INDEX embeddableIndex FOR (n:Document) ON (n.embedding) OPTIONS {indexConfig: {\vector.dimensions`: 1536, `vector.similarity_function`: 'cosine'}}`. The index requires the APOC plugin.

What happens if I don’t provide an OpenAI API key?

If OPENAI_API_KEY is not set, the server falls back to the all-MiniLM-L6-v2 sentence‑transformers model for generating embeddings.

What are the runtime dependencies?

Python 3.8+, uv package manager, Neo4j 5.0+ with the APOC plugin, and the Python packages fastmcp, neo4j, openai, python-dotenv, sentence-transformers, and pydantic.

Can I use this server without Neo4j?

No, a running Neo4j database with a vector index is required for both storing node embeddings and performing similarity searches.