MongoDB adds automated Voyage embeddings in Atlas Vector Search

MongoDB announced automated Voyage embeddings in Atlas Vector Search to improve retrieval accuracy and lower production costs and said MongoDB 8.3 is generally available.

By adding automated Voyage embeddings into the MongoDB stack, the company argued that it is removing the synchronization tax that occurs when enterprises have to stitch together separate vector databases, embedding services and operational stores.

The big picture is that MongoDB is positioning itself as a unified data platform that embeds, remembers, retrieves, protects and performs to deliver agentic AI. MongoDB is focusing on automated retrieval via integrated embeddings, vector search and re-ranking and persistent agent memory.

MongoDB CEO CJ Desai said the data layer is the hardest part of running AI agents. "To trust an agent at scale,
it has to retrieve the right context, hold memory across sessions, and operate at machine speed," said Desai.

• MongoDB: Q4 strong, but outlook light amid exec departures
• MongoDB integrates Voyage 4 models across its platform

MongoDB acquired Voyage AI for its reranking models in 2025.

The rollout, announced at MongoDB.local London, also included the following:

• MongoDB 8.3 general availability with a focus on AI mission critical workloads. The upgrade includes about 35% more throughput, 45% more read throughput and an expansion of native query expressions. The company said the goal with MongoDB 8.3 was to keep data transformation and prep inside the database.

• Atlas integration with Feast, an open-source feature store. The upshot is that Atlas can now act as a single backend for model training feature storage and real-time inference feature serving. MongoDB said Atlas integration with Feast reduces database sprawl, model drift between training and serving and overhead for machine learning teams.
• Long-term memory for JavaScript agents and general availability of LangChain JS Long-Term Memory Store integrated with MongoDB.
• Cross-Region Connectivity Support for AWS PrivateLink. App traffic across regions in MongoDB Atlas will be entirely on AWS' private backbone. The move creates one auditable and globally consistent security model.
• New MongoDB AI Skill badges for automatic Voyage embeddings, agentic memory and semantic search.