The next-generation database developer, which is holding its MongoDB.local NYC developer conference this week, has also established an alliance with startup Cognition to provide partners and customers with tools to migrate legacy database-based applications to modern data stacks.
Next-generation database developer MongoDB Tuesday debuted MongoDB 9.0, a major new release of the company’s core database platform, and Atlas Infinite, a new deployment option with “extreme elasticity” for the company’s popular MongoDB Atlas cloud database.
MongoDB, whose executives are holding an investor day today at the Nasdaq MarketSite in New York City, also launched Atlas Agent Engine, a new unified execution, memory and governance layer for production AI agents.
And the company announced a major alliance with startup Cognition, developer of the Devin “autonomous AI software engineer,” in a move to provide partners and customers with a faster way to shift aging code and data from legacy databases to MongoDB.
[Related: MongoDB CEO Desai Leaves For New Position At Meta, Database Developer Rehires Past CEO Ittycheria]
The whirlwind of activity, which also includes the company’s MongoDB.local NYC developer conference on Wednesday, follows the surprise news Monday that Chirantan “CJ” Desai had stepped down as MongoDB president and CEO—positions he assumed in early November 2025—to take a new job at Meta Group.
MongoDB 9.0 is the next generation of the company’s database engine, the core of the MongoDB Atlas fully managed cloud data and the foundation for the company’s Enterprise Advanced self-managed database for on-premises and private cloud deployments and the free Community Edition.
“Since our inception, we’ve never taken our foot off the gas when it comes to database innovation,” said Ben Cefalo, MongoDB chief product officer, core products, during a press briefing prior to today’s announcements.
Cefalo said the decision by MongoDB’s founders to originally develop the company’s database using a non-relational document database model and JSON data format was prescient.
“They saw that the real world doesn’t fit neatly into rigid tables,” he said, referring to relational databases. A document architecture, with its ability to handle evolving data schema and semi-structured and unstructured data, is generally seen as better suited for today’s AI applications.
“As we move deeper into the enterprise, the applications we powered grew more demanding, and we matched that growth with a constant drumbeat of innovation, staying ahead of our customers’ needs so they could keep building on us at scale,” he said.
MongoDB 9.0
The new 9.0 release, now generally available, offers major gains in performance, scalability, availability and security to meet the demands of modern AI applications. The release provides up to 35 percent faster find-one queries compared to MongoDB 8.0 (released two years ago), up to 30 percent faster update-one queries, up to 2x throughput on large instances, and up to 20 percent higher throughput for transactional workloads, according to the company.
Cefalo said MongoDB 9.0 provides enhanced industry-first and industry-only queryable encryption capabilities targeted for applications that process “highly sensitive data.”
MongoDB is calling the new Atlas Infinite, currently in public preview, “the largest architectural innovation” to the MongoDB Atlas cloud database since its 2016 launch.
Atlas Infinite separates compute and storage functions, allowing each to scale independently and providing a higher level of elasticity, responsiveness and retrieval accuracy for processing workloads—including the read/write workload activity spikes AI agents often generate—on real-time operational data.
MongoDB is renaming the current Atlas edition “Atlas Core” with Atlas Infinite a new deployment option. Internal tests using MongoDB 9.0 showed Atlas Infinite provided 189 percent greater throughput per dollar than Atlas Core on input/output-heavy workloads, according to the company. Atlas Infinite requires no application code changes and runs the same drivers, APIs and operational tooling as Atlas Core.
“Extreme elasticity and tunable performance at scale is what we hear customers need and expect from our database and it’s exactly what Atlas Infinite delivers,” Cefalo said.
Atlas Agent Engine
MongoDB also unveiled Atlas Agent Engine to help build, deploy, scale and govern production-grade AI agents. It provides a unified execution, memory, and governance layer that overcomes the data retrieval, persistent memory, and security and governance hurdles that often hinder efforts to move agentic applications from proof-of-concept testing into production.
“Everybody agrees that the potential of the AI and agentic era is obvious. But the implementation … feels like an iceberg—all this promise on the surface [and] massive complexity that lurks just underneath,” said Chirag Shah, MongoDB senior vice president of product, during the press conference. “The part where agentic projects struggle or die is when it’s time to take them to production.”
Agentic AI systems today often handle identity, audit, guardrails, and cost controls as separate systems. Atlas Agent Engine provides a single control plane to log every action by agents and humans and ensures governance policies are applied. The product leverages the embedding and reranking models technology MongoDB acquired when it bought Voyage AI in February 2025.
Atlas Agent Engine runs on MongoDB 9.0 and Atlas Infinite and is currently available in public preview.
Shah, in response to a question from CRN during the press conference, said MongoDB is working closely with its systems integrator partners with the new technologies—especially with Atlas Agent Engine because of the opportunities it provides as they work with clients to develop, deploy and manage complex agentic AI systems.
“Atlas Agent Engine brings the enterprise-ready capabilities, context and constraints needed to help AI agents deliver real-world impact,” said Ram Ramalingam, global lead of software engineering and head of RDE (reinvention deployed engineering) at Accenture, in an emailed response to a CRN inquiry.
“Combined with Accenture’s governance, architecture, and deep industry expertise, it creates a powerful foundation for accelerating AI transformation and delivering outcomes at scale. Our shared commitment to delivery and customer success makes this partnership particularly strong,” Ramalingam said.
The database company is also working to integrate Atlas Agent Engine with products offered by technology partners, including leading AI frontier models.
Cognition Alliance
MongoDB also announced that it has struck an alliance with Cognition, the startup developer of Devin, an autonomous AI software engineering tool, in a move to make it easier for businesses and organizations to migrate applications off legacy databases and onto MongoDB.
“As we look to customers, enterprises, who want to move to AI, their data is, oftentimes, extremely locked in,” said Rohit Batra, MongoDB senior vice president of product management, in an interview with CRN. “They are not able to access that data because [legacy] databases were not built for the scale and the use cases that AI typically solves for. And addressing that becomes challenging.”
Batra said such application modernization initiatives can be painstaking and expensive, updating the old applications’ business logic and code and then moving data from old supporting databases to new databases such as MongoDB.
MongoDB has been addressing the data migration issue with purpose-built deterministic tools to move data from relational databases with relational schemas to MongoDB’s database and its document architecture, Batra said.
Cognition’s Devin tackles the application redevelopment problem by automating software coding tasks. Devin can autonomously plan an application architecture, write code, install dependencies, run test suites, identify bugs and deploy production-ready applications.
Today the two companies launched Devin for MongoDB Modernizations, which connects Devin directly to MongoDB’s Application Modernization Platform (AMP). Connecting Devin to AMP’s tooling, which moves and validates data into the Atlas database, provides a faster, more automated way to move from legacy code to new production applications running on Atlas, the companies said.
The MongoDB-Cognition combination can significantly reduce migration timelines and free up engineering teams to build applications on MongoDB Atlas instead of working around legacy code.
“The entire application can now be moved onto a new [database] model stack,” Batra said. The two companies have also developed a playbook with system modernization instructions and use cases and is working with some early adopter customers.
As for solution providers, service providers and systems integrators working with the MongoDB-Cognition combination, “modernization is a huge opportunity,” Batra said, noting that many service providers are already working with Cognition as part of their system modernization practices. And partners with business transformation and vertical industry expertise can bring those capabilities to software migration projects.
Cognition already works with many of the same global and regional systems integrators that MongoDB does, said Art Levy, Cognition vice president of global partnerships, in the same interview with CRN.
Levy said MongoDB and Cognition have aligned their sales, product and partnership teams to support the Devin for MongoDB Modernization offering. “That’s what’s really setting this partnership up for success,” he said.
