MongoDB adds AI retrieval tools for live Atlas data

MongoDB has introduced new AI agent and retrieval tools in its Atlas platform, aimed at giving developers direct access to live operational data.

The additions include automated embeddings in Atlas using Voyage AI models, a new Atlas Embedding and Reranking API, a code retrieval model called voyage-code-4, vector search in Atlas Stream Processing, and a managed MCP server for connecting coding agents to Atlas.

The move addresses a common problem for teams building AI applications: linking agents and retrieval systems to operational databases without maintaining separate embedding pipelines, vector stores, and synchronisation processes. The new products bring retrieval and operational data handling into the same platform.

With automated embeddings, Atlas creates embeddings and updates indexes as documents are written or changed. This removes the need for developers to move data out of operational systems and keep separate vector databases aligned with source records.

The Atlas Embedding and Reranking API gives developers direct access to Voyage AI embedding and reranking models through Atlas, including for applications that do not run on MongoDB. voyage-code-4 is built specifically for code retrieval, with a focus on coding agents rather than general-purpose search.

MongoDB is also extending vector search to streaming data through Atlas Stream Processing. This is intended to let agents retrieve information from live event flows as well as stored data.

Customer examples

The Financial Times is using MongoDB's automated embeddings with Voyage AI models to improve semantic search across its journalism. MongoDB said the publisher had consolidated search workloads on its platform after managing search across multiple teams and products.

"Our job is to make the FT's journalism fast and easy to reach, however our readers come to it. With Automated Embedding and Voyage AI models on Atlas, we've improved retrieval accuracy while keeping costs in check across more than 100,000 searches a day, and being able to test and balance models lets us tune quality against cost as we go. With less infrastructure to run, the team can spend more of its time on the reading experience our subscribers rely on," said Elitsa Pavlova, Principal Engineer, CM Platform, at Financial Times.

Legal AI company Eve is using the Atlas Embedding and Reranking API to improve the relevance of material surfaced during a case.

"In legal AI, retrieval quality is foundational-the right evidence has to surface at the right moment. MongoDB's Atlas Embedding and Reranking API gives us a promising way to improve relevance directly in the RAG layer, while simplifying the infrastructure needed to build and evolve these experiences," said Urvesh Patel, Staff AI Engineer, at Eve.

Agent connections

Alongside the retrieval products, MongoDB has launched Atlas Managed MCP Server, a hosted service that connects AI agents to Atlas without requiring users to run their own server infrastructure. The service is designed for tools including Claude Code, Codex, Grok Build, and Devin.

Developers can also connect the data platform to ChatGPT, Claude, and Grok for querying data in plain language, and to development environments such as Cursor. MongoDB said the MCP server is already recording more than 30,000 installs a week.

MongoDB's pitch is that developers increasingly want AI tools to work directly with production data and schemas rather than with copies or isolated test environments. By using the same credentials and access controls already in Atlas, teams can manage how agents interact with data from a single place, according to the company.

"Too many organisations are running AI in production with an operational database, a vector store, a search engine, and embedding and reranking models, all from different vendors, bolted together instead of built for it," said Jim Scharf, Chief Technology Officer, at MongoDB.

"That's where stale data and errors creep in, and it's usually where teams spend their time babysitting instead of building. Agents raise the bar. They need to retrieve live context continuously and cannot wait on overnight batch jobs. MongoDB was built as an operational platform from the start, so retrieval and memory run on the same live data, nothing to sync, and agents act on what's happening instantly," said Scharf.

Pablo Stern-Plaza, Chief Product Officer, AI and Emerging Products, said MongoDB wants to be present across the software tools developers now use.

"The AI tools teams reach for keep changing, so our approach is to make sure MongoDB is present in all of them, whether a team is working in Claude or Codex, or running an agent in production. More of that building is now done by agents, and neither the agent nor the developer has to stop and set up a connection, so applications come together faster," said Stern-Plaza.

Partners also framed the launch around access to live application data. OpenAI highlighted MongoDB's plugin in ChatGPT, while Cognition pointed to the value of connecting Devin to production context.

"We're in the golden age of software engineering. The scope of what one engineer can build has exploded, and the unlock is agents working with real context," said Russell Kaplan, President, at Cognition, the company behind Devin.

"By connecting Devin to MongoDB Atlas, engineers can hand off well-scoped tasks knowing Devin is working from live application data, not stale assumptions, and spend their own time on the harder problems," said Kaplan.