What is Vector database?

A vector database is built to store and query embeddings, the numeric vectors that an embedding model produces to capture the meaning of text, images, or other data. Instead of matching exact keywords, it answers nearest-neighbor queries: given a query vector, return the stored vectors closest to it in high-dimensional space, which corresponds to the items most similar in meaning. To do this quickly over millions of vectors, these systems use approximate nearest-neighbor indexes (such as HNSW) and add the operational features applications need, metadata filtering, namespaces, and scalable updates. Vector databases are the backbone of semantic search and RAG: you embed your documents once, store them, and at query time embed the user's question and retrieve the closest chunks to ground the model's answer. Several are exposed as MCP servers so an agent can store and search embeddings directly through tool calls. Dedicated stores like Pinecone, Qdrant, and Chroma specialize in this, while general databases increasingly add vector support, for example Postgres via pgvector.