Vector MCP Server

Overview

Vector MCP Server is a standardized collection management system across vector database technologies, supporting ChromaDB, PGVector, Couchbase, Qdrant, and MongoDB. It allows AI Agents to perform hybrid search for document information (lexical/vector), create collections with documents stored on the local filesystem or URLs, and utilize collections for retrieval augmented generation (RAG).

Features

• Multi-Database Support: Single interface for ChromaDB, Couchbase, MongoDB, Qdrant, and PGVector
• Hybrid Search: Combine lexical and vector search for optimal retrieval
• Collection Management: Create and manage document collections across different vector databases
• Document Sources: Support for local filesystem and URL-based documents
• RAG Implementation: Built for retrieval augmented generation use cases
• Standardized Interface: Consistent API across different vector database backends

Supported Vector Databases

• ChromaDB: In-memory and persistent vector storage
• Couchbase: Enterprise NoSQL with vector search capabilities
• MongoDB: Document database with Atlas Vector Search
• Qdrant: High-performance vector similarity search engine
• PGVector: PostgreSQL extension for vector embeddings

Use Cases

• Building RAG applications with multiple vector database backends
• Semantic search across document collections
• AI agents with long-term memory capabilities
• Multi-modal search combining text and metadata
• Cross-database vector search evaluation

Technical Implementation

Heavily inspired by the RAG implementation of Microsoft's Autogen V1 framework, providing a battle-tested approach to vector-based retrieval.

Pricing

Free and open-source. Individual database hosting costs may apply.