MongoDB

Database · AI Infrastructure · Developer Tools

Launch

MongoDB brings automated embeddings and managed MCP server to Atlas

The update folds embedding pipelines and agent connectivity straight into the database, undercutting bolt-on vector search rivals.

• MongoDB announced at its MongoDB.local Build Fest event that Automated Embedding in Atlas powered by Voyage AI is now generally available, alongside a new Atlas Managed MCP Server.
• MongoDB is the developer database behind Atlas, its cloud platform; these launches extend Atlas to serve as the data and memory layer for AI coding agents.
• The Managed MCP Server is fully hosted with one-click OAuth, letting agents like Claude Code, Codex, Grok Build and Devin query, inspect and update Atlas data with no connection strings or infrastructure to run.
• Automated Embedding removes the need for teams to build and sync separate embedding pipelines; MongoDB also shipped a standalone Embedding and Reranking API, the voyage-code-4 model, and vector search in Atlas Stream Processing.
• MongoDB's existing MCP server already logs more than 30,000 npm installs per week, and its Voyage AI embedding models currently top the Retrieval Embedding Benchmark (RTEB).
• The Financial Times said Automated Embedding let it consolidate splintered search tools while improving semantic search accuracy across more than 100,000 daily searches and holding down retrieval costs.
• By pulling embedding, retrieval and agent connectivity inside the operational database, MongoDB is betting it can beat point-solution vector databases on both accuracy and operational simplicity for production AI agents.