Weaviate homepage

Weaviate

by Weaviate B.V.

Open-source AI database for vector search, hybrid search, RAG, and agent memory

Agent PlatformAssistant

Last reviewed 2026-06-20

Visit websiteDocsAlternatives

Weaviate is an open-source, AI-native vector database that stores both objects and their vector embeddings (numeric representations of text, images, and other data) so developers can run semantic search, keyword (BM25) search, and hybrid combinations of the two with structured filtering, at scale and with cloud-native fault tolerance. Written in Go and released under the BSD-3-Clause license, it can vectorize data automatically through integrated modules for OpenAI, Cohere, Hugging Face, Google, and other model providers, and exposes Python, Go, TypeScript/JavaScript SDKs plus GraphQL and REST APIs. It is the retrieval and memory layer many RAG (retrieval-augmented generation) and AI-agent applications are built on. Beyond the core database, Weaviate has added higher-level products: a Query Agent that turns natural-language questions into optimized database queries, integrated Embeddings, an Engram managed-memory service for agents, and a built-in MCP (Model Context Protocol) server preview so LLM clients and coding agents can query the database directly. Weaviate is aimed primarily at developers and engineering teams. The company was founded in 2019 by Bob van Luijt, Etienne Dilocker, and Micha Verhagen, is headquartered in Amsterdam, Netherlands, and reports more than 20 million open-source downloads.

What it can do

  • Open-source vector storage and search

    Assistant

    Stores objects together with their vector embeddings and searches them at scale, with cloud-native fault tolerance and horizontal scaling; the core engine is open source under the BSD-3-Clause license.

    source

  • Hybrid search with keyword and metadata filtering

    Assistant

    Combines vector (semantic) search with traditional keyword (BM25) search and structured filtering, plus multi-tenancy for isolating data per customer.

    source

  • Integrated vectorization (embeddings)

    Assistant

    Generates vectors from text, images, and other data automatically through built-in modules for providers such as OpenAI, Cohere, Hugging Face, and Google, so no external embedding pipeline is required.

    source

  • Query Agent (agentic search)

    Supervised

    A Weaviate Cloud service that takes a natural-language question and automatically decides which collections to search and which filters, sorts, and search types to apply, returning either a natural-language answer (Ask mode) or raw filtered objects (Search mode). The user submits each query, so it is supervised rather than autonomous.

    source

  • Engram managed memory for agents

    Assistant

    A managed memory service, built on Weaviate, described as long-term memory for AI agents; positioned as a memory layer rather than an agent that plans or acts on its own.

    source

  • Built-in MCP server

    Supervised

    Weaviate ships an MCP (Model Context Protocol) server, introduced as a preview in v1.37.0 (April 2026), so MCP-compatible LLM clients and coding agents can query the database directly without a custom integration layer.

    source

Strengths

  • +Open-source core (BSD-3-Clause) that can be fully self-hosted, with managed cloud and Bring Your Own Cloud (BYOC) options for teams that want control over data and cost
  • +Strong hybrid search (vector plus BM25 keyword) with metadata filtering and multi-tenancy out of the box
  • +Integrated vectorizers and a built-in MCP server reduce the glue code needed to wire embeddings and agents to the database

Limitations

  • −Usage-based cloud pricing (charged on vector dimensions and storage) plus monthly minimums can be hard to predict for growing workloads
  • −Self-hosting the open-source database means operating and scaling search infrastructure yourself
  • −It is retrieval and memory infrastructure, not an autonomous agent; planning and orchestration live in the application built on top

Overview

Weaviate is an open-source, AI-native vector database: it stores objects together with their embeddings (numeric representations of text, images, and other data) and lets developers search them by meaning, by keyword, or a hybrid of both, with structured filtering and cloud-native scalability. It is the retrieval and memory layer many semantic-search, recommendation, and RAG (retrieval-augmented generation) systems are built on. The engine is written in Go and released under the BSD-3-Clause license. Weaviate was founded in 2019 by Bob van Luijt, Etienne Dilocker, and Micha Verhagen, is based in Amsterdam, and reports more than 20 million open-source downloads and thousands of customers including names such as Booking and Intuit (vendor-stated).

What it does

The core product stores and indexes high-dimensional vectors and supports vector search, keyword (BM25) search, and hybrid search that blends the two, plus metadata filtering and multi-tenancy. Integrated vectorizer modules generate embeddings automatically from providers such as OpenAI, Cohere, Hugging Face, and Google, so teams can skip a separate embedding pipeline. On top of the database sit higher-level products: a Query Agent that translates a natural-language question into an optimized query (deciding which collections, filters, and search types to use) and returns either a natural-language answer or raw objects; integrated Embeddings; an Engram managed-memory service positioned as long-term memory for agents; and a built-in MCP (Model Context Protocol) server (preview from v1.37.0, April 2026) so MCP-compatible clients and coding agents can query the database directly. Weaviate is infrastructure, not an agent: the database answers retrieval queries and does not plan or act on its own, so its baseline autonomy is assistant-level, with the Query Agent and MCP server adding supervised agentic access where a human still submits each request.

Integrations & setup

Weaviate exposes GraphQL and REST APIs plus SDKs for Python, Go, and TypeScript/JavaScript. It integrates with the common RAG orchestration frameworks (LangChain, LlamaIndex, Haystack, Agno) and with embedding/model providers (OpenAI, Cohere, Hugging Face, Google) through its module system. Deployment options span fully self-hosted (open-source), managed Weaviate Cloud (Shared and Dedicated), and Bring Your Own Cloud (BYOC) that runs in the customer's own AWS, Google Cloud, or Azure account.

Pricing

Freemium. The open-source database is free to self-host. Weaviate Cloud has an always-free tier (one cluster, 100,000 objects, capped memory and storage, community support). Paid managed tiers reported in 2026 are Flex (pay-as-you-go, month-to-month, from about $45/month, 99.5% SLA), Plus (prepaid annual, from about $280/month, adds SSO/SAML and 30-day backups, 99.9% SLA), and Premium (from about $400/month, choice of shared or dedicated deployment, 99.95% SLA, dedicated technical account team). Paid usage is charged on vector dimensions and storage and varies by cloud provider and region. The platform is SOC 2 Type II certified, with HIPAA compliance available on Enterprise/Dedicated (AWS). Weaviate moved to this Flex/Plus/Premium model in late 2025, replacing the older Serverless/Enterprise tiers.

Best for / not for

Best for developers and engineering teams that want an open-source vector database they can self-host or run as a managed service, with strong hybrid search and flexible deployment (self-hosted, managed cloud, or BYOC). It suits teams that value control over data location and cost, or that want integrated vectorization and a built-in MCP server. Less suited to teams that want a fully managed, zero-ops service with no self-hosting decisions, or to those expecting an out-of-the-box autonomous agent: Weaviate is retrieval and memory infrastructure, and the agent intelligence lives in the application built on top.

Alternatives

Pinecone, Qdrant, Chroma, and Milvus are the main vector-database alternatives. Pinecone is fully managed and serverless; Qdrant and Milvus are open-source databases with managed clouds, like Weaviate; Chroma is a lightweight open-source store popular for prototyping. LangChain and LlamaIndex are not competitors but the orchestration frameworks most commonly used alongside Weaviate.

What people are saying

We aggregate real LinkedIn discussion into sentiment for the agents people search most. Weaviate isn't tracked yet, want it added? Request tracking.

FAQ

Is Weaviate an AI agent?+

No. Weaviate is an open-source vector database and retrieval platform. It supplies the knowledge and memory layer (semantic search, hybrid search, RAG, and agent retrieval via its Query Agent) that AI agents and applications call, but the database itself does not plan or take actions. Its core function is best described as assistant-level retrieval infrastructure; the Query Agent adds supervised agentic search on top, where a human submits each query.

What is Weaviate used for?+

Storing and searching vector embeddings to power semantic search, hybrid (vector plus keyword) search, recommendations, and retrieval-augmented generation (RAG). Developers use it to give LLM applications and agents fast, filtered access to domain-specific or up-to-date data, and increasingly as managed memory for agents.

Is Weaviate open source and can I self-host it?+

Yes. The Weaviate database is open source under the BSD-3-Clause license and is written in Go, so it can be self-hosted on your own infrastructure. Weaviate also offers a managed Weaviate Cloud (Shared and Dedicated) and a Bring Your Own Cloud (BYOC) option that runs in the customer's own cloud account.

How much does Weaviate cost?+

Weaviate is freemium. The open-source database is free to self-host, and Weaviate Cloud has an always-free tier (one cluster, capped objects and storage). Paid managed tiers reported in 2026 are Flex (pay-as-you-go from about $45/month), Plus (prepaid annual from about $280/month, with SSO/SAML and a 99.9% SLA), and Premium (from about $400/month, with shared or dedicated deployment and a 99.95% SLA). Paid usage is charged on vector dimensions and storage, and Enterprise/Dedicated contracts are custom.

Sources

Last reviewed 2026-06-20

Alternatives & related

Pinecone
Managed vector database for semantic search, RAG, and AI agent memory
LangChain
Open-source framework and platform for building and deploying LLM agents
LlamaIndex
Open-source data framework for RAG pipelines and data-grounded agents