Azure Cosmos DB

Azure Cosmos DB is a fully managed, cloud-native NoSQL database using a schema-agnostic JSON document model, designed for ultra-low latency and elastic scale globally, with software development kits for Python, Node.js, Java and .NET.

Global distribution as a first-class property rather than a configuration option is the defining characteristic. Most databases can be replicated across regions with enough work; Cosmos DB is built so that multi-region operation is the normal case, which matters for an application serving users on several continents where the alternative is either poor latency for most of them or a distributed system the team maintains itself.

Unifying transactions, search and embeddings in one service is the current positioning and it addresses a real architectural annoyance. An AI application typically needs an operational database, a search index and a vector store, which means three systems to keep consistent and three failure modes. Built-in vector and hybrid search on the operational data removes two of them, which is why Microsoft positions Cosmos DB for retrieval augmented generation and agentic applications.

One naming change is worth knowing before evaluating: Azure DocumentDB is the new name for vCore-based Azure Cosmos DB for MongoDB, a fully managed MongoDB-compatible database built on the open source DocumentDB engine under the Linux Foundation. Automatic indexing, integration with Microsoft Foundry and enterprise service level agreements complete the offer. Billing is by request units and storage through the Azure pricing calculator rather than as plan tiers.