What it is: We have launched the MongoDB VFS (Virtual Files System) for LangChain Deep Agents, a backend integration of the LangChain Deep Agents file system protocol. With this integration, MongoDB can now serve as a storage backend for LangChain Deep Agents, providing AI agents with a durable, searchable workspace that persists across sessions, deployments, and sub-agents. More specifically, it implements Deep Agents' BackendProtocol, routing grep, glob, and Is through MongoDB (vector search + full-text search + hybrid $rankFusion) while forwarding all other file operations (read, write,edit, upload_files, download_files) directly to S3.

Who it's for: Developers building and deploying Deep Agents. LangChain's built-in backends are ephemeral or single-node, so anything long-running or multi-agent requires hand-rolling persistence. This integration removes that plumbing entirely, giving them durable, searchable agent memory via the same read_file / glob / grepcalls they already write, so they can stay focused on agent design rather than storage infrastructure.

Why it matters: This integration gives Deep Agents durable, production-grade memory: retrieval runs through MongoDB's vector, full-text, and hybrid search; file storage rides on S3; and both remain accessible across sessions and sub-agent handoffs, all through the same filesystem calls agents already make. Developers can now ship long-running, multi-agent systems without having to build or manage the persistence layer themselves.

How to get started:  Check out the Official Guide to start building with Deep Agents, or head over to the LangChain documentation to learn more about it.