Managed Deep Agents 0.8 Launches with User-Scoped Memory, HTTP Channels, and Parallel Web Search

LangChain has launched Managed Deep Agents (MDA) v0.8, an update to its production agent platform designed to simplify infrastructure for authentication, memory, and channel integration. The release addresses key production challenges by adding support for user-owned credentials and user-level memory, which allows agents to store caller-specific context separately from shared agent knowledge to prevent context leakage. MDA v0.8 also introduces HTTP channels, enabling integration with internal tools and customer portals via webhooks, complementing existing Slack support which now includes file handling. Additionally, the platform now includes built-in web search powered by Parallel, which is free to use during the beta period. Developers can configure these features through declaration files, with default policies restricting user memory in shared contexts to ensure privacy.