Pinecone is a fully managed vector database made by Pinecone Systems, Inc. A vector database stores embeddings, lists of numbers that represent the meaning of a text or an image, and finds the stored items whose numbers are closest to those of a query. This is how an AI application finds the passages that match a question, for example in RAG (retrieval-augmented generation, where found passages are added to a prompt).
A program writes records into an index through the API and queries the same index. Each index can serve full-text search, semantic search on dense vectors, sparse-vector search and hybrid search. With integrated embedding, Pinecone turns text into vectors on its own servers. Results can be narrowed with metadata filters and then reranked, and namespaces keep each customer's data apart inside one index. New writes become searchable within seconds, and Pinecone selects and tunes the index algorithms itself.
Pinecone is a hosted service, also sold through the AWS, Google Cloud and Microsoft cloud marketplaces. A free Starter plan has usage limits; paid plans are a flat monthly fee or pay per use with a monthly minimum.