Azure Cosmos DB

Combines vector search and full-text search (BM25) to deliver context-aware and keyword-relevant results within the same document model

Integrate with Microsoft Foundry and Foundry Tools to store agent conversations and model interactions securely with enterprise control and observability

Generate queries by asking questions using natural language for building and using agents and multi-agent apps

Integrate both NoSQL and vector data into retrieval-augmented generation (RAG) workloads and AI agents

Built-in vector search powered by DiskANN for storing and querying vectors and NoSQL data efficiently in a serverless vector database

Enjoy the best end-to-end Azure Cosmos DB experience with the native document API

Mirror Azure Cosmos DB data into Microsoft Fabric to enable near-real-time analytics, AI, and reporting without ETL processes

Avoid overprovisioning and reduce expenses by letting Azure Cosmos DB adjust throughput dynamically with autoscale, maintaining performance during spikes

Build faster with flexible JSON schema and seamless data model evolution with schema-free JSON and automatic indexing

Build without capacity planning using serverless in Azure Cosmos DB, allowing you to scale and pay only for what you use

Code with favorite SDKs including Python, Node.js, Java, .NET, Go, and more with full Azure Cosmos DB SDK support

Build event-driven apps using Azure Cosmos DB change feed to process events in real time with Azure Functions, AKS, and Azure Container Apps

Manage multitenant workloads at scale with Azure Cosmos DB fleets, organizing tenant accounts, leveraging shared throughput pools, and gaining fleet-wide analytics

Build highly responsive apps with less than 10ms latency SLA and up to 99.999% availability across multiple regions

Accelerate queries with Global Secondary Indexes by creating read-only containers with different partition keys, turning cross-partition queries into fast single-partition lookups

Get started with a free tier option to test and develop applications

Choose from consumption-based serverless options, provisioned throughput, traditional scale-up of nodes, and scale-out architectures for automatic elasticity

Maintain high availability for single-write region accounts by automatically failing over at the partition level, minimizing disruption and eliminating manual intervention

Scale out automatically with automatic partitioning as data volume and throughput demand grow without impacting availability or requiring manual intervention

Benefit from embedded security with 34,000 full-time equivalent engineers dedicated to security initiatives, partnerships with 15,000 specialized security experts, and over 100 compliance certifications