MongoDB announced Atlas Agent Engine, Atlas Infinite and MongoDB 9.0 ahead of its MongoDB.local conference in New York. Atlas Agent Engine and Atlas Infinite are in public preview; MongoDB 9.0 is expected to become generally available soon. MongoDB says the three launches position it as a platform for AI agents and data, after months of private previews with core customers. The company expects the platform to help move AI agents into production faster and boost growth.

Atlas Agent Engine is a governed platform for building and running AI agents, and is available to all Atlas customers in public preview without an additional contract or purchase. Its memory system includes semantic, episodic, taxonomic and procedural memory. Tool calls receive least-privilege authorization enforced when an action occurs. Retrieval uses MongoDB’s Voyage AI embedding and reranking models, while customers can use the memory and governance layers separately or alongside the runtime. The system supports all models and frameworks, and MongoDB says it has no vendor lock-in: switching from a frontier model to an open-weight model should require only a configuration change. MongoDB uses the engine to run Holly, its internal agent, and plans an on-premises option before general availability as customer demand grows for on-premises deployments to manage token budgets. The company has also lined up systems integrators to support broader agent deployments.

Atlas Infinite, also in public preview, is a new Atlas deployment option that separates computing from storage to scale for bursty workloads, including agents that may be idle before needing substantial resources. MongoDB reports 189% more performance per dollar on I/O-heavy workloads and pricing based on actual compute and storage use. Its existing Atlas offering will be renamed Atlas Core. MongoDB says Atlas Infinite can also handle Postgres workloads.

MongoDB 9.0 powers both launches and is expected to deliver 20% higher transaction throughput and 30% faster updates, with efficiency gains particularly for larger instances. Compared with MongoDB 8.0, it offers up to 35% faster find-one queries, up to 30% faster update-one queries and up to 2x throughput for large instances. It also adds Queryable Encryption support for prefix, suffix and substring searches on encrypted data.

MongoDB announced Atlas Agent Engine, Atlas Infinite and MongoDB 9.0 ahead of its MongoDB.local conference in New York. Atlas Agent Engine and Atlas Infinite are in public preview; MongoDB 9.0 is expected to become generally available soon. MongoDB says the three launches position it as a platform for AI agents and data, after months of private previews with core customers. The company expects the platform to help move AI agents into production faster and boost growth.

Atlas Agent Engine is a governed platform for building and running AI agents, and is available to all Atlas customers in public preview without an additional contract or purchase. Its memory system includes semantic, episodic, taxonomic and procedural memory. Tool calls receive least-privilege authorization enforced when an action occurs. Retrieval uses MongoDB’s Voyage AI embedding and reranking models, while customers can use the memory and governance layers separately or alongside the runtime. The system supports all models and frameworks, and MongoDB says it has no vendor lock-in: switching from a frontier model to an open-weight model should require only a configuration change. MongoDB uses the engine to run Holly, its internal agent, and plans an on-premises option before general availability as customer demand grows for on-premises deployments to manage token budgets. The company has also lined up systems integrators to support broader agent deployments.

Atlas Infinite, also in public preview, is a new Atlas deployment option that separates computing from storage to scale for bursty workloads, including agents that may be idle before needing substantial resources. MongoDB reports 189% more performance per dollar on I/O-heavy workloads and pricing based on actual compute and storage use. Its existing Atlas offering will be renamed Atlas Core. MongoDB says Atlas Infinite can also handle Postgres workloads.

MongoDB 9.0 powers both launches and is expected to deliver 20% higher transaction throughput and 30% faster updates, with efficiency gains particularly for larger instances. Compared with MongoDB 8.0, it offers up to 35% faster find-one queries, up to 30% faster update-one queries and up to 2x throughput for large instances. It also adds Queryable Encryption support for prefix, suffix and substring searches on encrypted data.