# Vector databases and retrieval

Where an agent looks things up in its own data: vector databases, search engines with vector and keyword search, embedding and reranking models, and toolkits for retrieval-augmented generation.

78 tools, in alphabetical order. Nothing is ranked by popularity.

- [Amazon S3 Vectors](https://modelmoot.com/tools/amazon-s3-vectors.md): An Amazon S3 feature that stores vectors in vector buckets and runs similarity queries on them. (Amazon Web Services; not open source, hosted service)
- [Annoy](https://modelmoot.com/tools/annoy.md): A C++ library with Python bindings for approximate nearest-neighbour search using memory-mapped read-only indexes. (Spotify; open source)
- [Apache Solr](https://modelmoot.com/tools/apache-solr.md): A search platform built on Lucene with an HTTP API, faceting and dense vector search. (Apache Software Foundation; open source)
- [Azure AI Search](https://modelmoot.com/tools/azure-ai-search.md): A managed Azure search service with vector, keyword, hybrid and semantic ranking, with agentic retrieval features. (Microsoft; not open source, hosted service)
- [Chroma](https://modelmoot.com/tools/chroma.md): A database for embeddings and documents that runs in-process, as a server, or as a hosted service. (Chroma; open source, hosted service)
- [Chroma MCP server](https://modelmoot.com/tools/chroma-mcp.md): Chroma's MCP server for creating collections and storing and querying documents in Chroma. (Chroma; open source)
- [Cloudflare Vectorize](https://modelmoot.com/tools/cloudflare-vectorize.md): A vector database on Cloudflare that is bound to Workers and queried from them or over a REST API. (Cloudflare; not open source, hosted service)
- [Cognee](https://modelmoot.com/tools/cognee.md): An open source engine that turns documents and data into a knowledge graph with vectors for agent memory. (Cognee; open source, hosted service)
- [Cohere Embed and Rerank](https://modelmoot.com/tools/cohere.md): Hosted embedding and reranking models from Cohere, reached through its API for retrieval pipelines. (Cohere; not open source, hosted service)
- [ColBERT](https://modelmoot.com/tools/colbert.md): The reference implementation of ColBERT, a late-interaction neural retrieval model, from Stanford. (Stanford Future Data Systems; open source)
- [DiskANN](https://modelmoot.com/tools/diskann.md): A graph-based approximate nearest-neighbour library, from Microsoft, that serves large indexes from SSD with little memory. (Microsoft; open source)
- [Elasticsearch](https://modelmoot.com/tools/elasticsearch.md): A distributed search and analytics engine with full-text, vector and hybrid search, and a REST API. (Elastic; open source, hosted service)
- [Faiss](https://modelmoot.com/tools/faiss.md): A C++ library with Python bindings for similarity search and clustering of dense vectors. (Meta; open source)
- [FalkorDB](https://modelmoot.com/tools/falkordb.md): A graph database that runs as a Redis module, queried with Cypher and with vector index support, aimed at GraphRAG. (FalkorDB; not open source, hosted service)
- [FlagEmbedding](https://modelmoot.com/tools/flagembedding.md): A toolkit and model family, including the BGE embedding and reranker models, for retrieval. (Beijing Academy of Artificial Intelligence; open source)
- [Graphiti](https://modelmoot.com/tools/graphiti.md): A Python library for building knowledge graphs that change over time, made for agent memory. (Zep AI; open source)
- [GraphRAG](https://modelmoot.com/tools/graphrag.md): A Microsoft data pipeline and query system that builds a knowledge graph with community summaries from text for retrieval. (Microsoft; open source)
- [Haystack](https://modelmoot.com/tools/haystack.md): A Python framework from deepset for retrieval and agent pipelines assembled from components. (deepset; open source, hosted service)
- [hnswlib](https://modelmoot.com/tools/hnswlib.md): A header-only C++ library with Python bindings that implements HNSW approximate nearest-neighbour search. (nmslib; open source)
- [Hugging Face MCP Server](https://modelmoot.com/tools/hugging-face.md): Hugging Face's MCP server for searching models, datasets, papers and Spaces on the Hub. (Hugging Face; open source, hosted service)
- [Infinity](https://modelmoot.com/tools/infinity.md): A database for hybrid search that combines dense vectors, sparse vectors, tensors and full-text search. (InfiniFlow; open source)
- [Jina AI](https://modelmoot.com/tools/jina-ai.md): Hosted embedding, reranking and reader APIs, plus small open models for search over text, images and long documents. (Jina AI; not open source, hosted service)
- [LanceDB](https://modelmoot.com/tools/lancedb.md): An embedded retrieval library and database built on the Lance columnar format, usable locally or in the cloud. (LanceDB; open source, hosted service)
- [LangChain](https://modelmoot.com/tools/langchain.md): A library of model, tool and retrieval integrations with a ready-made agent interface, in Python and JavaScript. (LangChain; open source)
- [LangChain4j](https://modelmoot.com/tools/langchain4j.md): A Java library for model applications with a common API for providers, vector stores, tools and agents. (LangChain4j community; open source)
- [LightRAG](https://modelmoot.com/tools/lightrag.md): A Python library for retrieval-augmented generation that builds a knowledge graph from documents and queries both graph and vectors. (HKUDS; open source)
- [LlamaIndex](https://modelmoot.com/tools/llamaindex.md): A Python framework for connecting models to data, with retrieval pipelines, agents and event-driven workflows. (LlamaIndex; open source, hosted service)
- [Marqo](https://modelmoot.com/tools/marqo.md): A search engine and hosted service that generates embeddings itself and handles text and image search. (Marqo; open source, hosted service)
- [MCP Toolbox for Databases](https://modelmoot.com/tools/mcp-toolbox.md): Google's open-source MCP server that connects agents to databases such as BigQuery, Cloud SQL, Spanner and Postgres. (Google; open source)
- [Meilisearch](https://modelmoot.com/tools/meilisearch.md): A search engine with a REST API that combines keyword and semantic search through hybrid queries. (Meilisearch; open source, hosted service)
- [Meilisearch MCP server](https://modelmoot.com/tools/meilisearch-mcp.md): Meilisearch's MCP server for managing indexes, documents and settings and running searches. (Meilisearch; open source)
- [Memgraph](https://modelmoot.com/tools/memgraph.md): An in-memory graph database compatible with Cypher and the Bolt protocol, with vector search and GraphRAG tooling. (Memgraph; not open source, hosted service)
- [Memvid](https://modelmoot.com/tools/memvid.md): A memory layer that packs an agent data, index and metadata into a single portable file. (Memvid; open source)
- [Milvus](https://modelmoot.com/tools/milvus.md): A distributed vector database that runs as a single process, in a cluster, or as the Zilliz Cloud service. (Zilliz; open source, hosted service)
- [Milvus MCP server](https://modelmoot.com/tools/milvus-mcp.md): Zilliz's MCP server for searching and managing collections in a Milvus vector database. (Zilliz; open source)
- [Mixedbread](https://modelmoot.com/tools/mixedbread.md): A hosted search service and embedding and reranking models, with a store for multimodal documents. (Mixedbread; not open source, hosted service)
- [MongoDB Atlas Vector Search](https://modelmoot.com/tools/mongodb-atlas-vector-search.md): A vector search feature of MongoDB Atlas that indexes embeddings stored in documents alongside other fields. (MongoDB; not open source, hosted service)
- [Neo4j](https://modelmoot.com/tools/neo4j.md): A graph database queried with Cypher that also has vector indexes, used for GraphRAG and agent memory. (Neo4j; open source, hosted service)
- [Neo4j MCP servers](https://modelmoot.com/tools/neo4j-mcp.md): Neo4j Labs MCP servers for running Cypher queries and for building a memory graph in Neo4j. (Neo4j Labs; open source)
- [Nomic Embed](https://modelmoot.com/tools/nomic-embed.md): Open-weight embedding models and a hosted API from Nomic, together with the Atlas tool for exploring embeddings. (Nomic; not open source, hosted service)
- [OpenSearch](https://modelmoot.com/tools/opensearch.md): A search and analytics engine, forked from Elasticsearch, with a k-NN vector search plugin and neural search. (OpenSearch Software Foundation; open source, hosted service)
- [OpenSearch MCP server](https://modelmoot.com/tools/opensearch-mcp.md): OpenSearch project's MCP server for listing indices and running searches on an OpenSearch cluster. (OpenSearch project; open source)
- [OpenViking](https://modelmoot.com/tools/openviking.md): A context database for agents that holds memory, knowledge and skills in a virtual file system. (Volcengine; open source)
- [ParadeDB](https://modelmoot.com/tools/paradedb.md): A Postgres extension and distribution that adds BM25 full-text search and analytics, used with pgvector for hybrid search. (ParadeDB; open source, hosted service)
- [Paragon](https://modelmoot.com/tools/paragon.md): Integration infrastructure for agents and SaaS products, with an ActionKit for acting in users' apps and managed sync. (Paragon; not open source, hosted service)
- [pgvector](https://modelmoot.com/tools/pgvector.md): A PostgreSQL extension that adds a vector type and nearest-neighbour search with HNSW and IVFFlat indexes. (pgvector; open source)
- [pgvectorscale](https://modelmoot.com/tools/pgvectorscale.md): A PostgreSQL extension that adds a disk-based DiskANN-style index and compression for use with pgvector. (Timescale; open source)
- [Pinecone](https://modelmoot.com/tools/pinecone.md): A managed vector database offered only as a cloud service, with serverless indexes. (Pinecone; not open source, hosted service)
- [Pinecone MCP server](https://modelmoot.com/tools/pinecone-mcp.md): Pinecone's MCP server for searching documentation and managing and querying Pinecone indexes. (Pinecone; open source)
- [Qdrant](https://modelmoot.com/tools/qdrant.md): A vector database and search engine written in Rust, with filtering on payload fields. (Qdrant; open source, hosted service)
- [Qdrant MCP server](https://modelmoot.com/tools/qdrant-mcp.md): Qdrant's MCP server that stores and finds information in a Qdrant vector collection. (Qdrant; open source)
- [R2R](https://modelmoot.com/tools/r2r.md): A retrieval system with a REST API that provides ingestion, hybrid search, knowledge graphs and agentic retrieval. (SciPhi; open source, hosted service)
- [RAG-Anything](https://modelmoot.com/tools/rag-anything.md): A framework built on LightRAG that handles text, images, tables and equations in documents for retrieval. (HKUDS; open source)
- [Ragas](https://modelmoot.com/tools/ragas.md): A Python library for evaluating LLM applications, with metrics for retrieval, answer quality and agent behaviour. (Vibrant Labs; open source)
- [RAGatouille](https://modelmoot.com/tools/ragatouille.md): A Python library that makes late-interaction ColBERT retrieval simple to train and use. (Answer.AI; open source)
- [RAGFlow](https://modelmoot.com/tools/ragflow.md): A retrieval-augmented generation engine with deep document parsing, chunking and an agent workflow editor. (InfiniFlow; open source, hosted service)
- [Redis](https://modelmoot.com/tools/redis-query-engine.md): An in-memory data store whose query engine adds vector similarity, full-text and secondary indexing. (Redis; open source, hosted service)
- [Redis Agent Memory Server](https://modelmoot.com/tools/redis-agent-memory-server.md): A server from Redis that gives agents working memory for a session and long-term memory with search. (Redis; open source)
- [Redis MCP Server](https://modelmoot.com/tools/redis.md): Redis's own MCP server for reading and writing strings, hashes, lists, sets, streams and search indexes. (Redis; open source)
- [rerankers](https://modelmoot.com/tools/rerankers.md): A Python library with one interface for many reranking models and APIs. (Answer.AI; open source)
- [Sentence Transformers](https://modelmoot.com/tools/sentence-transformers.md): A Python library for using and training embedding, reranker and sparse encoder models. (Hugging Face; open source)
- [Spring AI](https://modelmoot.com/tools/spring-ai.md): A Spring project for Java applications that call models, use tools, retrieve data and talk to MCP servers. (Spring (Broadcom); open source)
- [sqlite-vec](https://modelmoot.com/tools/sqlite-vec.md): A SQLite extension for storing vectors and running nearest-neighbour queries in SQL. (Alex Garcia; open source)
- [Tantivy](https://modelmoot.com/tools/tantivy.md): A full-text search engine library written in Rust, inspired by Lucene. (Quickwit; open source)
- [Text Embeddings Inference](https://modelmoot.com/tools/text-embeddings-inference.md): A server from Hugging Face for serving embedding, reranking and sequence classification models. (Hugging Face; open source)
- [turbopuffer](https://modelmoot.com/tools/turbopuffer.md): A hosted search database that keeps data on object storage and caches it, with vector and full-text search. (turbopuffer; not open source, hosted service)
- [Typesense](https://modelmoot.com/tools/typesense.md): A search engine with typo tolerance, filtering and vector search, with built-in embedding generation. (Typesense; open source, hosted service)
- [Upstash Vector](https://modelmoot.com/tools/upstash-vector.md): A serverless vector database with an HTTP and REST API, built for use from edge and serverless functions. (Upstash; not open source, hosted service)
- [USearch](https://modelmoot.com/tools/usearch.md): A single-file similarity search engine for vectors with bindings for many languages. (Unum; open source)
- [Vearch](https://modelmoot.com/tools/vearch.md): A distributed vector search system that stores documents with vectors and offers hybrid retrieval. (Vearch; open source)
- [Vectara](https://modelmoot.com/tools/vectara.md): A hosted retrieval-augmented generation platform with an API for indexing documents, search and grounded answers. (Vectara; not open source, hosted service)
- [VectorChord](https://modelmoot.com/tools/vectorchord.md): A PostgreSQL extension for vector search that uses IVF with RaBitQ quantization and supports disk-friendly indexes. (TensorChord; open source)
- [Vertex AI Vector Search](https://modelmoot.com/tools/vertex-ai-vector-search.md): A managed Google Cloud vector search service based on the ScaNN algorithm, with hybrid search. (Google Cloud; not open source, hosted service)
- [Vespa](https://modelmoot.com/tools/vespa.md): A search and recommendation engine that combines vector, text and structured search with on-node model inference. (Vespa.ai; open source, hosted service)
- [Voyage AI](https://modelmoot.com/tools/voyage-ai.md): Hosted embedding and reranking models for retrieval, with domain variants such as code and finance. (Voyage AI; not open source, hosted service)
- [Weaviate](https://modelmoot.com/tools/weaviate.md): A vector database that stores objects with vectors and combines vector, keyword and hybrid search. (Weaviate; open source, hosted service)
- [Zep](https://modelmoot.com/tools/zep.md): A hosted context and memory service that builds a temporal knowledge graph from agent conversations and business data. (Zep AI; not open source, hosted service)
- [Zvec](https://modelmoot.com/tools/zvec.md): An in-process vector database from Alibaba that runs inside an application with no separate server. (Alibaba; open source)

HTML: https://modelmoot.com/tools/c/retrieval. All categories: https://modelmoot.com/tools.md
