What is the difference between vector and hybrid search?

Vector search ranks results by embedding similarity, so it matches on meaning even when the words differ. Hybrid search combines that with keyword (BM25) matching, which recovers exact terms like a product code or a name that a vector alone can underrank. Weaviate and Milvus expose hybrid; Chroma and Pinecone here lean on vectors and filters.

Do I need to create the embeddings myself?

Depends on the server. Pinecone's search-records runs the embedding from your text query for you. Others expect the collection to already hold vectors you embedded at index time. Match the tool to how your data was loaded.

MCP servers that can query a vector database

5 verified servers expose a tool that can query a vector database

Semantic search runs on a vector database: text is embedded into vectors, and a query returns the nearest matches by meaning rather than exact words. When an agent needs that kind of retrieval, the vector-store servers expose a query tool over your collection.

These verified servers let an agent query a vector database for the closest matches to a piece of text.

Top pick

Pinecone

Official

Pinecone's official developer MCP server: search indexes, manage records, rerank results, and look up Pinecone docs from your agent.

vector-search

Tool:

search-records

search-records queries an index from a plain text query, running the embedding for you and offering metadata filters and reranking, the managed option that hides the vector mechanics.

Pick 2

Weaviate

Official

Weaviate's built-in MCP server: hybrid vector and keyword search, schema inspection, and object upserts against a Weaviate vector database.

vector-search

Tool:

weaviate-query-hybrid

weaviate-query-hybrid blends vector similarity with BM25 keyword matching in one call, which recovers exact terms a pure vector search would rank too low.

Pick 3

Milvus

Zilliz

Official

Zilliz's official Milvus MCP server: vector, full-text, and hybrid search plus collection management over a Milvus vector database.

vector-search

Tools:

milvus_vector_search
milvus_text_search

milvus_vector_search runs dense-vector similarity with selectable distance metrics, and milvus_text_search adds full-text BM25, for teams running Milvus at scale.

Pick 4

Chroma

Official

Chroma's official MCP server: manage collections and run semantic, metadata, and full-text search over a Chroma vector database.

vector-search

Tool:

chroma_query_documents

chroma_query_documents does semantic search with filtering over a Chroma collection, a common pick for a local or lightweight vector store.

Pick 5

LanceDB

Adiom

Community

A maintained MCP server for agentic RAG over a local LanceDB index: hybrid search across a document catalog and its chunks.

vector-search79

Tool:

chunks_search

chunks_search finds the relevant chunks for a document in a LanceDB catalog, suited to retrieval over embedded document chunks.

What to know

The query tools differ in how they match. A pure vector search ranks by embedding similarity. A hybrid search blends that with keyword (BM25) matching, which catches exact terms a vector alone can miss. Some servers run the embedding for you from a text query; others expect you to pass vectors. The choice usually follows your data: an agent retrieves from whichever store the documents were indexed into, so the server tracks the collection rather than the other way around.

This is retrieval, not memory in itself. A vector store answers what is similar to this query right now; it does not on its own track what an agent learned across sessions or decide what is worth keeping. Building that durable, shared layer is a separate problem from the nearest-neighbor lookup these tools perform.

Questions

What is the difference between vector and hybrid search?: Vector search ranks results by embedding similarity, so it matches on meaning even when the words differ. Hybrid search combines that with keyword (BM25) matching, which recovers exact terms like a product code or a name that a vector alone can underrank. Weaviate and Milvus expose hybrid; Chroma and Pinecone here lean on vectors and filters.
Do I need to create the embeddings myself?: Depends on the server. Pinecone's search-records runs the embedding from your text query for you. Others expect the collection to already hold vectors you embedded at index time. Match the tool to how your data was loaded.