MongoDB Expands AI Retrieval Tools for On-Premises and Private Cloud Deployments

MongoDB unveiled a suite of artificial intelligence (AI) capabilities aimed at helping enterprises address two of the biggest obstacles to deploying production AI applications: retrieval accuracy and regulatory compliance. Announced at MongoDB.local Bengaluru on June 30, the updates expand the company's AI retrieval stack beyond MongoDB Atlas, enabling organizations to deploy the same capabilities across on-premises, private cloud, hybrid, and local environments.

The release introduces Voyage Context 4, Hybrid Search, and Native Reranking while making MongoDB Search and Vector Search generally available for MongoDB Enterprise Advanced and Community Edition. According to the company, the additions enable organizations to build retrieval-augmented generation (RAG) applications without relying on separate search systems or redesigning applications as they move between deployment environments. 

MongoDB Targets AI Retrieval Accuracy

The announcement comes as enterprises continue investing in enterprise AI infrastructure to move AI workloads from experimentation into production. While foundation models continue to improve, many organizations now face challenges retrieving accurate, up-to-date enterprise data that large language models (LLMs) require to generate reliable responses.

Among the new capabilities, Native Reranking is available in public preview for MongoDB Atlas. Powered by Voyage AI models, it reranks search results inside the database, eliminating the need for external APIs or separate reranking services. MongoDB said the feature improves retrieval quality by up to 30% based on Voyage AI's benchmarking.

Voyage Context 4, now generally available, is an embedding model designed for long documents. Rather than processing isolated chunks of text, it preserves document-level context to improve retrieval quality. Hybrid Search, also generally available, combines full-text and vector search within a single database query, allowing AI applications to retrieve information from live operational data instead of separate indexed copies.

"The biggest barrier to enterprise AI in production and at scale isn't the LLM. It's memory, retrieval, accuracy, and compliance," Ben Cefalo, Chief Product Officer, Core Products, MongoDB, said in a statement.

The launch reflects a broader shift across enterprise AI toward improving context for AI applications. Vendors increasingly view retrieval as a critical layer for reducing hallucinations and improving the reliability of agentic AI systems. That trend aligns with broader efforts around moving AI projects from pilot to production, where infrastructure decisions increasingly focus on trusted enterprise data rather than model performance alone.

Search Moves Beyond MongoDB Atlas

MongoDB also expanded Search and Vector Search to MongoDB Enterprise Advanced, allowing organizations to deploy AI retrieval capabilities behind their own firewalls while maintaining the same APIs and development experience available in Atlas. The company said more than 20 large banks and financial institutions evaluated the capabilities before their general availability.

The move targets organizations operating under strict data residency, sovereignty, and compliance requirements, complementing broader investment in sovereign AI infrastructure. MongoDB also made Search and Vector Search generally available in Community Edition, allowing developers to prototype AI applications locally before scaling to Enterprise Advanced or Atlas without switching databases or rearchitecting applications.

The announcements build on MongoDB's acquisition of Voyage AI, which brought embedding and reranking models into the company's platform as it expands its AI capabilities beyond its origins as a NoSQL database vendor.

Separately, MongoDB announced plans to train 2 million developers in India by 2030 through partnerships with the All India Council for Technical Education (AICTE), HCL GUVI, and ICT Academy of Kerala. The company also launched the Bengaluru to the Bay startup challenge, which offers winning founders $50,000 in MongoDB Atlas credits, travel support, and access to MongoDB.local San Francisco.

As enterprises continue integrating AI into business applications, MongoDB is positioning its database platform as both a system for operational data and a retrieval layer designed to provide AI models with current, enterprise-specific context across cloud and self-managed environments.