Weaviate review: Open-source vector database with hybrid search, RAG and agents for AI developers

What Weaviate does

Weaviate is an open-source AI database aimed at developers building search, retrieval-augmented generation and agent applications. It stores and searches high-dimensional vectors, supports hybrid search that mixes vector and BM25 keyword matching, and offers tenant isolation, index compression and configurable backups. Around the core database sit a Query Agent, which turns natural-language questions into database queries, built-in embedding generation, and Engram, a service for personalized experiences that adapt to each user.

You can run Weaviate yourself, use it as a managed service in Weaviate Cloud, or deploy it as a Kubernetes package in your own VPC. Developers can work through SDKs for Python, Go, TypeScript or JavaScript, or through GraphQL and REST APIs.

Pricing on Weaviate Cloud has a free plan at $0 per month with no credit card required. It allows 100,000 objects, 1 GB memory, 10 GB disk, one collection and up to 3 tenants, with Embeddings capped at 2,000 requests per day and Query Agent at 1,000 requests per month. The Flex plan is pay-as-you-go and starts at $45 per month on a shared cloud cluster. Premium is a prepaid contract, shared or dedicated, with pricing available through sales; the pricing page says its minimum starts from $400 per month. Vector dimensions, storage and backups are billed by usage, and the pricing page shows no annual-billing prices.

• Vector database for high-dimensional vectors
• Hybrid vector and BM25 keyword search
• Query Agent for natural-language queries
• Built-in embedding generation
• Engram for adaptive personalized experiences
• Native multi-tenancy with tenant isolation
• Vector index compression
• Self-hosted, managed cloud or Kubernetes in your VPC

Our take

Weaviate fits developers and teams who want a vector database they can start on for free and later self-host or move to a managed cloud. The free plan is small, so it suits exploration rather than production, and the cap of one collection and three tenants will be reached quickly by multi-tenant products. Flex adds replication and higher limits from $45 per month, but SSO/SAML, phone and Slack support only appear from Premium, which requires talking to sales. Costs depend on vector dimensions, storage and backups, so a bill needs estimating with the pricing calculator rather than reading a flat price. Teams wanting a simple per-seat price will find a different model here.