The vector database that keeps your data encrypted at rest, in transit, and in use.
Similarity search directly on encrypted embeddings — self-hosted in your VPC or on-prem. AES-256-GCM with your KMS key, per-operation key derivation, keys zeroized after use.
CyborgDB is an encrypted vector database — similarity search directly on encrypted embeddings, with per-operation key derivation via your KMS and keys zeroized after use. Only the final candidates are decrypted. It runs self-hosted and standalone — persisting to local disk (default) or S3 — exposes a familiar API that slots into existing AI pipelines, and can be GPU-accelerated with NVIDIA cuVS.
<5ms
p95 encrypted query latency
100M+
vectors per index
Zero
data leaves your boundary
Choose Your Path
Learn the Fundamentals
Understand the architecture and principles behind encrypted vector search
Start Building
Get hands-on with CyborgDB in minutes using our quickstart guide
Explore Deployment Options
Discover embedded libraries, managed service, and custom deployment options
Documentation Structure
Introduction
Core concepts, deployment models, and getting started guides
CyborgDB Service
CyborgDB self-hosted with REST API and client SDKs
CyborgDB Embedded
Self-hosted, embedded deployments with Python and C++ bindings
Integrations
LangChain and other framework integrations for seamless adoption
How to Use These Docs
Navigation tips and conventions used throughout these docs
