MongoDB launches Atlas Agent Engine, Atlas Infinite as MongoDB 9.0 goes GA

MongoDB launched Atlas Agent Engine, a governed agentic platform with a sophisticated memory architecture, Atlas Infinite, an Atlas deployment option that separates compute and storage for better scaling, and MongoDB 9.0, which will be generally available soon.

The trio, announced ahead of its MongoDB.local conference in New York, represents MongoDB's move to be a broad platform for AI agents and data. MongoDB's latest products have been in private preview with core customers for months and the bet is the company can significantly boost its growth by being a data platform that moves AI agents into production faster.

MongoDB's headliner is Atlas Agent Engine, which is in public preview, but all three announcements build on each other. Chirag Shah, senior vice president, general manager and head of AI platform at MongoDB, said Atlas Agent Engine is "an open, secure and a governed agentic infrastructure" where "memory is a foundation not an afterthought."

The news from Mongo lands a day after CJ Desai left the company to lead Meta Enterprise Platform. MongoDB's Dev Ittycheria is back in CEO fold after stepping aside less than a year ago and will oversee a product launch that can juice growth at the company.

Atlas Agent Engine features semantic, episodic, taxonomic and procedural memory. Every tool call is authorized with least privilege and enforced upon action not audited after the fact.

MongoDB, which is holding an investor day too, is also betting that its neutral status also plays well with developers and enterprises. "There's zero vendor lock-in," said Shah, who added that moving from a frontier model to an open weight model should be "as simple as a configuration change."

All Atlas customers can use Atlas Agent Engine in public preview and there's no additional contract or purchase since MongoDB is consumption based. MongoDB has been using Atlas Agent Engine to run Holly, its internal agent.

• MongoDB delivers strong Q2, outlook
• MongoDB launches Voyage AI, search tools aimed at on-prem, private cloud deployments
• MongoDB adds automated Voyage embeddings in Atlas Vector Search

Shah added that Atlas Agent Engine will be available as an on-premises option before generally availability. MongoDB is seeing a big surge in on-premises demand as customers look to save on the token budget.

Key points about Atlas Agent Engine include:

• Retrieval is powered by MongoDB's Voyage AI embedding and reranking models.
• Customers can adopt the memory and governance layers independently or with the runtime.
• The system can use all models and frameworks.
• MongoDB has lined up a bevy of systems integrators for broader AI agent deployments.

Atlas Infinite is MongoDB's other big launch that's in public preview. Atlas Infinite is a new deployment option available to Atlas customers. Pablo Stern-Plaza, MongoDB's chief product officer of AI and emerging products, said Atlas Infinite is the company's "biggest architectural innovation since its inception."

By separating compute and storage, Stern-Plaza said Atlas Infinite will be better suited to scale and handle the burst traffic that AI agents can require. Agents can often sit idle for long stretches and then require massive demand for compute and storage.

According to MongoDB, Atlas Infinite can provide 189% more performance per dollar on I/O heavy workloads. In addition, Atlas Infinite has a new pricing model tied to actual compute and storage consumption.

MongoDB's current Atlas offering will be rebranded Atlas Core.

Ben Cefalo, MongoDB's chief product officer for core products, said Atlas Infinite can take Postgres workloads.

The core powering Atlas Agent Engine and Atlas Infinite is MongoDB 9.0, which has 20% higher transaction throughput and 30% faster updates adding up to more efficiency, especially for larger instances. MongoDB 9.0 also includes Queryable Encryption with support for prefix, suffix, and substring searches on encrypted data.

Relative to MongoDB 8.0, MongoDB 9.0 features:

• Up to 35% faster find-one queries.
• Up to 30% faster update-one queries.
• Up to 2x throughput for large instances.

Also see:

• Constellation ShortList™ Hybrid-Cloud and MultiCloud NoSQL Databases
• Constellation ShortList™ Next-Gen Databases: RDBMS for On-Premises