# GraphANN

> GraphANN is a commercial, closed-source graph-based vector search engine. It navigates a graph over your document corpus and recomputes embeddings only for visited nodes, so raw vectors are never persisted at rest — up to 95%+ less storage than traditional vector databases at scale, with built-in RAG retrieval, hard multi-tenant isolation, and fully air-gapped, single-binary deployment.

## Product

- [Overview](https://graphann.com/): GraphANN is a graph-based vector search engine and vector database alternative. Sub-linear similarity search, up to 95% less embedding and index storage than traditional vector databases, built-in RAG retrieval, and air-gap deployment. No embeddings at rest.
- [GraphANN vs Pinecone, Weaviate & Qdrant](https://graphann.com/compare): Side-by-side vector database comparison: GraphANN against Pinecone, Weaviate, Qdrant, and pgvector for similarity search. Sub-linear queries, up to 95% less storage, no embedding persistence. Honest about the gaps.
- [Cost calculator (storage & query TCO)](https://graphann.com/tco): Estimate your total cost of ownership for vector search at scale. GraphANN's graph-based approach typically delivers roughly 5x to 14x less storage than traditional vector databases like Pinecone, Weaviate, or Qdrant.
- [Case studies](https://graphann.com/case-studies): Real GraphANN deployments: embedding-model hot-swap without reindexing, air-gapped production, RAG retrieval at ~50ms p50, multi-tenant SaaS isolation. Graph-based vector search under real-world load.
- [One-page summary](https://graphann.com/leaflet): GraphANN at a glance: sub-linear queries, up to 95% less storage, air-gap deployment. The vector search trade-off most databases don't talk about.

## Documentation

- [Documentation index](https://graphann.com/docs): GraphANN documentation: getting started, HTTP and gRPC API reference, Kubernetes and air-gap deployment guides, multi-tenant vector search, and migration from Pinecone, Weaviate, or Qdrant.
- [Getting started](https://graphann.com/docs/getting-started): Deploy GraphANN with Docker, Kubernetes, or a single signed binary. No external vector database, no indexing pipeline, no embeddings at rest. Get similarity search running in minutes.
- [API reference (HTTP & gRPC)](https://graphann.com/docs/api-reference): GraphANN HTTP and gRPC API reference for inserting documents, running similarity search queries, managing tenants, and monitoring vector search performance.
- [Multi-tenant isolation](https://graphann.com/docs/multi-tenancy): GraphANN multi-tenancy: hard tenant isolation per vector index, per-tenant API keys, RBAC roles, and independent quotas. No query-time filtering overhead. Built for SaaS embedding products.
- [Deployment (on-prem, cloud, air-gap)](https://graphann.com/docs/deployment): Deploy GraphANN as a single binary, Docker Compose, Kubernetes StatefulSet, or fully managed service. Air-gapped and on-premises environments supported with zero outbound calls.
- [Migration from Pinecone, Weaviate, Qdrant](https://graphann.com/docs/migration): Migration guide from Pinecone, Weaviate, Qdrant, or pgvector to GraphANN. Typical migration runs in under an hour with no downtime and no re-embedding of your source corpus.

## Company

- [Contact](https://graphann.com/contact): Talk to the GraphANN team about graph-based vector search for your workload. We reply within one business day.
- [Privacy policy](https://graphann.com/privacy): How GraphANN collects, uses, and protects your data. Air-gap, on-prem, and cloud deployments covered.