pgvector is an open-source PostgreSQL extension for storing embeddings and running vector similarity search with SQL. It is best for teams already using PostgreSQL that want semantic search, RAG retrieval, or AI memory without operating a separate vector database, while accepting PostgreSQL scaling and tuning tradeoffs.
pgvector is an open-source PostgreSQL extension that stores embeddings and enables vector similarity search using SQL.
pgvector is a $0 MIT-licensed PostgreSQL extension for storing embeddings, querying vector similarity with SQL, and keeping RAG, semantic search, recommendations, and AI memory data inside an existing PostgreSQL database instead of adopting a separate vector database, with paid cost coming from PostgreSQL hosting, compute, memory, storage, backups, monitoring, and operations. It adds vector-oriented data types, distance operators, functions, and approximate nearest-neighbor indexing options to PostgreSQL, so developers can store embeddings beside relational records and filter results with normal SQL predicates, joins, roles, transactions, and backup workflows.
Verifiable product facts are central to evaluating pgvector. It is implemented as a PostgreSQL extension. Its public repository is hosted at github.com/pgvector/pgvector. The project uses the MIT license. It supports exact nearest-neighbor search and approximate indexing. The visible feature set includes HNSW indexes and IVFFlat indexes. It supports common vector distance patterns such as L2 distance, inner product, cosine distance, and L1 distance. It also exposes vector-oriented types beyond the standard vector type, including half-precision, binary, and sparse vector options where supported by the installed version. pgvector is accessed through PostgreSQL clients because its API surface is SQL, not a separate hosted service API.
The main value is architectural simplicity for PostgreSQL-centered teams. A product catalog, document table, support-ticket store, or agent memory table can keep embeddings and metadata in one transactional database. That makes pgvector especially useful when retrieval needs tenant filters, permissions, timestamps, joins, full-text search, or existing operational controls. It also reduces integration work for teams already using PostgreSQL migrations, connection pooling, database roles, backup policies, and observability.
The tradeoff is that pgvector inherits PostgreSQL's operational boundaries. Query latency, recall, index build time, write throughput, memory use, vacuum behavior, and backup size all depend on schema design, index choice, hardware, PostgreSQL configuration, table size, filter selectivity, update rate, and hosting limits. Dedicated vector databases may be stronger for very large distributed collections, managed multi-region retrieval, vector-first APIs, or workloads where vector search must be isolated from transactional database traffic.
Pricing should be read in two layers. The pgvector software itself is free and open source, with no license fee. Production use still has real cost: PostgreSQL hosting, larger instances for memory-heavy indexes, storage for embeddings and indexes, backup retention, monitoring, failover, staff time, and managed-provider support. Managed PostgreSQL providers that support pgvector publish their own plan prices; for example, common entry paid plans in this category include Supabase Pro at $25 per month per organization and Neon Launch at $19 per month, while self-hosted PostgreSQL can start at $0 software cost but requires infrastructure and administration.
Was this helpful?
pgvector is a practical choice for PostgreSQL-centered teams that need vector search without adding a separate retrieval service. It is strongest for SQL-heavy applications, RAG systems tied to relational data, and moderate-scale semantic search. Teams should benchmark before using it for very large, distributed, or latency-critical vector workloads.
Adds vector data types and operators directly to PostgreSQL.
Supports approximate search indexes such as HNSW and IVFFlat, depending on version and configuration.
Enables vector search alongside filters, joins, and ordering in SQL.
Stores embeddings with related application data under PostgreSQL transaction semantics.
Works through normal PostgreSQL clients and application frameworks.
Software licensing is free; infrastructure and operations costs depend on the PostgreSQL environment.
Uses PostgreSQL security controls where configured, but does not independently provide compliance certification.
Can combine vector similarity with SQL filters and PostgreSQL text-search patterns.
$0
Ready to get started with pgvector?
View Pricing Options →pgvector works with these platforms and services:
We believe in transparent reviews. Here's what pgvector doesn't handle well:
Weekly insights on the latest AI tools, features, and trends delivered to your inbox.
No specific 2026 release-note claim is included here. Check the official pgvector GitHub releases for current version changes.
Vector Database
Fully managed vector database for RAG and AI search with serverless storage, hybrid sparse-dense indexes, integrated embedding and rerank models, and managed retrieval workflows.
Vector Database
Open-source AI-native vector and hybrid search database with built-in modules for embedding, generative AI (RAG), reranking, and multimodal data — available self-hosted or as Weaviate Cloud.
Vector Database
Open-source, Rust-built vector similarity search engine with payload filtering, hybrid search, quantization, and a fully managed Qdrant Cloud — popular for RAG, recommendation, and agent memory.
AI Memory & Search
Milvus: Open-source vector database to analyze and search billions of vectors with millisecond latency at enterprise scale.
AI Infrastructure
Open-source, embedded multimodal vector database designed to live next to your AI app rather than as a separate service.
AI Search & Embeddings
Open-source AI search platform for large-scale RAG, personalization, and recommendation — battle-tested at Yahoo, with hybrid vector + lexical + structured ranking.
No reviews yet. Be the first to share your experience!
Get started with pgvector and see if it's the right fit for your needs.
Get Started →Take our 60-second quiz to get personalized tool recommendations
Find Your Perfect AI Stack →Explore 20 ready-to-deploy AI agent templates for sales, support, dev, research, and operations.
Browse Agent Templates →Everything builders need to know about vector databases — how they work under the hood, which one to choose (with real pricing and benchmarks), and how to implement them in RAG pipelines, agent memory systems, and multi-agent architectures.
AI agents without memory restart from zero every conversation, wasting time and money. Here's how the three types of agent memory work, why they matter for your business, and which tools actually deliver results in 2026.