// category · 11 tools · 8 open source
Vector Databases
Stores for embeddings and agent memory, managed or self hosted, powering retrieval, semantic search and long term recall.
A vector database stores embeddings, the lists of numbers a model produces from text or images, and finds the closest ones to a query in milliseconds. For agents it does two jobs: retrieval for RAG (fetching the right chunks of your documents) and long term memory (recalling what a user said or decided in earlier sessions). This category covers both the databases and the memory layers built on top of them. When the answer lives on the web rather than in your own documents, a search API fits better; see Tavily vs Exa vs Firecrawl.
Servers you run or rent. Qdrant (Rust, Apache 2.0) filters on JSON payloads during the search and combines sparse and dense vectors for hybrid search; see how to use Qdrant. Weaviate (Go) can create embeddings inside the database, keeps each tenant on its own shard and ships an MCP endpoint; see what is Weaviate. Milvus targets very large collections, and Pinecone and turbopuffer are managed services only.
Embedded and in database options. LanceDB and Chroma can run inside your process with data in a folder, which suits single agents and desktop apps; LanceDB vs Qdrant explains the trade off. If you already run Postgres, pgvector adds vector search to it without a new service.
Memory layers. Mem0, Zep and Graphiti sit above a store: they extract facts from conversations, merge duplicates and track when facts change. Compare them in AI agent memory: Mem0 vs Letta vs Zep.
How to choose for an agent: first, decide how you will isolate users. Per user filters or tenants must be set by your code, never by the model. Second, check hybrid search, because agents look up exact names, error codes and IDs that pure semantic search can miss. Third, check how the store reaches the agent: a framework integration, an MCP server, or your own tool. Fourth, compare the free tiers; for example, the Qdrant pricing page lists a free single node cluster and the Weaviate pricing page a free tier of 100,000 objects. For a full retrieval pipeline, see how to build a RAG pipeline with LangGraph.
Chromatrychroma.comDeveloper friendly open source embedding database that runs embedded, as a server or in Chroma Cloud.Vector DatabasesOpen source★ 29.6k
Weaviateweaviate.ioWeaviate is an open source vector database in Go with hybrid BM25 plus vector search, per user tenants, vectorizer modules and a built in MCP server.Vector DatabasesOpen source★ 16.9k
LanceDBlancedb.comEmbedded, multimodal vector database built on the Lance columnar format.Vector DatabasesOpen source★ 11.8k
Qdrantqdrant.techQdrant is an open source vector database in Rust (Apache 2.0) with payload filtering, BM25 plus dense hybrid search and an official MCP server.Vector DatabasesOpen source★ 35.3k
Milvusmilvus.ioCloud native vector database designed for billion scale similarity search, with a managed Zilliz service.Vector DatabasesOpen source★ 46.5k
Zepgetzep.comManaged memory and context platform for agents that turns chats and business data into a temporal knowledge graph, built on the open source Graphiti engine.Vector DatabasesFreemium★ 5k
Mem0mem0.aiMemory layer for AI agents that extracts and recalls user facts across sessions.Vector DatabasesOpen source★ 68k
Graphitihelp.getzep.comOpen source Python library from Zep for building temporal knowledge graphs that track when facts change, backed by Neo4j, FalkorDB or Amazon Neptune.Vector DatabasesOpen source★ 32.3k
pgvectorgithub.comPostgres extension for vector similarity search, letting you keep embeddings next to relational data.Vector DatabasesOpen source★ 23.5k
turbopufferturbopuffer.comServerless vector and full text search built on object storage for low cost search at large scale.Vector DatabasesPaidHN 7
Pineconepinecone.ioFully managed serverless vector database for similarity search at scale.Vector DatabasesFreemiumHN 1// guides
Guides to vector databases
Best Open Source Vector Database in 2026
The best open source vector database for agents and RAG in 2026: Qdrant, Milvus, pgvector, Chroma, LanceDB and Weaviate compared by license, setup and scale.
Qdrant Vector Database: How to Use It (2026 Guide)
Qdrant is an open source vector database in Rust. How to use it: run it, store text, filter per user and add BM25 hybrid search, tested on Qdrant 1.20.
What Is Weaviate? Features, Pricing and Use Cases
Weaviate is an open source vector database in Go with hybrid search and a built in MCP server. How it works, a tested agent memory example and cloud pricing.
LanceDB vs Qdrant: Which Should You Use?
LanceDB vs Qdrant for AI agents: an embedded library on files or a vector search server. Setup, per user filters, hybrid search, scaling and the pick.
AI Agent Memory: Mem0 vs Letta vs Zep (2026)
AI agent memory explained: how Mem0, Letta and Zep store and recall long term context, how they differ from LangGraph's built in memory, and which to choose.
LanceDB vs Milvus: Which Should You Use?
LanceDB vs Milvus for AI agents: an embedded library on files, or a vector database that grows from a local file to a cluster. Setup, limits and the pick.
How to Build a RAG Pipeline with LangGraph
How to build a RAG pipeline with LangGraph: index your docs, wire retrieve and generate nodes, add a grading loop with a retry cap, then turn it into an agent.
What Is Agent Memory in Claude Code? Explained
What is agent memory in Claude Code? Plain Markdown files: CLAUDE.md you write and auto memory Claude writes. Where it lives, its limits and how to edit it.
How to Use LangChain with Ollama (2026 Guide)
How to use LangChain with Ollama: install langchain-ollama, pick a tool calling model, raise the context window and build a local agent with tools and RAG.