Artificial Intelligence | MongoDB Blog

Artificial Intelligence
The Claude Agent SDK is a package that lets you embed Claude Code's own agent loop directly into your application: web search, file search, file editing, MCP server connections, and more, all without having to define those tools or manage the agent loop yourself. The SDK also automatically persists session transcripts to the local file system by default. Claude calls it the Agent SessionStore.
Building AI applications has never been easier. Making them reliable in production remains the real challenge.
Every team building AI agents discovers one fundamental truth: without the right information, an agent cannot reliably answer a question or take the correct action.
When a huge surge in demand meets limited resources, an awful experience usually isn’t far behind. Fans feel it when trying to compete for concert tickets with bots and scalpers. Diners feel it when they try to get a reservation at a trendy restaurant that’s fully booked for months. Websites and applications feel it when they go viral and get an accidental DDoS from a “Reddit Hug of Death.”
I've spent more than 25 years in technology, which means I'm rarely surprised by new tech—I usually see it up close before the public does. As an IT leader, I never evaluate a new capability in isolation; it's always weighed against real enterprise use cases, and against the risks that come with them. These days, I tell people that being a CIO in the age of AI means holding two jobs at once: the engineer reinforcing the dam, and the operator who has to open the floodgates anyway.
Some of you may have noticed a different, and perhaps familiar, name at the top of this post.
This is a guest blog post, written by Devin Pratt, Research Director at IDC.
Organizations across Greater China are heavily investing in data-driven technologies to deliver innovative digital services and experiences. According to Morgan Stanley Research, AI investments in China alone could deliver a 52% return on invested capital by 2030*.
In production AI, what the system retrieves shapes everything that follows. It determines whether an application surfaces the right context in the first place and how much irrelevant information gets passed to an LLM. That affects two things you care about most: answer quality and cost.
The token economy is here. And as soon as you have an economy, people start budgeting. So it’s no surprise that most conversations about AI now come back to the token budget. Teams are being asked to show the value of inference—the action phase of AI—spend and justify the cost. The answer traces back to the data that was retrieved and fed to the LLM. This step is intuitively referred to as “retrieval.”
The CTOs I talk to in regulated industries aren’t debating whether to build with AI. That decision has been made. What they're navigating is a harder question: how do you build AI at the standard your compliance framework requires, when most of the tooling was designed for environments you're not allowed to use?
There’s a common theme to the conversations I’ve been having with AI teams lately: change. Constant, head-spinning change. Teams across industries are evaluating and re-evaluating model providers, agent frameworks, and harnesses on a continuous basis. At MongoDB, we believe that your choice of technology partner—specifically, your data platform—should simplify how you build with AI. It should deliver performance at scale, enable you to build and run anywhere, and it should allow you to choose your own providers and frameworks. This is exactly what MongoDB offers, and it’s why more than 67,000 customers rely on us for their most important applications. The organizations seeing the most AI success are the ones whose technology stacks are set up for the current pace of change. For example, DevRev’s AgentOS platform is powered by MongoDB Atlas. AgentOS handles billions of requests each month, for everything from AI-assisted insights and analytics to internal communications and development. Relying on MongoDB Atlas has helped DevRev get innovations to market faster, and enables the company to scale seamlessly as it grows. MongoDB is ideal for agentic AI in two key ways. First, an agent is only as smart as its context—which requires blending short-term memory, long-term knowledge, and enterprise data. Because this information is highly dynamic and unstructured, JSON is the ideal format. It provides the schema flexibility inherently needed by the data and allows attaching metadata like IDs and confidence scores. MongoDB stores JSON natively and provides the scale and consistency required to run thousands of concurrent agents. Second, it’s designed for how agents work. As memory accumulates, agents must pinpoint the precise context needed for a request. MongoDB solves this by providing state-of-the-art information retrieval capabilities (search, vector search, hybrid search, embeddings) directly where the operational data already lives, eliminating the need to constantly sync data across separate systems. Customers get high-precision semantic retrieval without the operational headache of managing multiple fragmented products. A good example of how MongoDB powers agents is ElevenLabs. The company relies on MongoDB Atlas to power the long-term memory and knowledge base for its autonomous agents. By leveraging Atlas Search and Vector Search, ElevenLabs enables their agents to retain complex context and deliver highly personalized interactions in real-time. Adobe, meanwhile, chose MongoDB as the long-term memory and reasoning layer for Journey Agent, its composite multimodal AI agent that unifies Adobe's marketing suite and orchestrates end-to-end customer journeys. Adobe leverages MongoDB Atlas Search and Atlas Vector Search together to power the sub-100 millisecond hybrid search the agent needs to act in real time. Defining an open standard for agent memory Last month, MongoDB partner LangChain announced the launch of Context Hub in La