LangChain Releases Managed Deep Agents 0.8 with User Auth, Scoped Memory, and HTTP Channels
Summary
- • LangChain's MDA 0.8 adds user-owned credentials and user-scoped memory for production agents
- • HTTP channels let agents run in portals, support systems, and other product surfaces
- • Built-in web search via Parallel integration is now included out of the box
- • Release targets four core production challenges: memory, auth, channels, and tool management
Details
MDA 0.8 general release
Managed Deep Agents 0.8 ships user-owned auth, user-level memory, HTTP channels, Slack file transfer, and integrated web search
Identity-scoped authentication
User-owned credentials let agents act within per-user permission boundaries; auth scoping is configured via a declarative identity file in the project directory
User-level memory scoping
Memory is scoped to the authenticated caller, giving agents per-user context retention across sessions; complements existing durable agent-level memory
HTTP channel support
HTTP channels allow agents to be deployed in internal tools, customer portals, and support systems that can send webhooks — broadening deployment surfaces
Web search integration
Built-in search powered by Parallel is included out of the box in MDA 0.8, no additional configuration needed
Slack file transfer support
Added file transfer capability within Slack channel integrations, extending existing Slack support in Managed Deep Agents
Production infrastructure bundling
LangChain says building equivalent agent infra (auth, memory, channels, tools) typically takes engineering teams quarters of roadmap; MDA packages it as a managed service
Deep Agents harness foundation
MDA is built atop Deep Agents, LangChain's code-first framework for long-running, mission-critical agents; MDA adds managed infrastructure on top
LangChain Managed Deep Agents 0.8 feature breakdown and context
What This Means
LangChain's MDA 0.8 addresses the real friction of moving agents from prototype to production: who the agent can act as, what it remembers about each user, and how it connects to existing surfaces and tools. The user-scoped auth and memory model is a meaningful step toward agents that are both secure and contextually aware across multi-user enterprise deployments. By pre-bundling these capabilities, LangChain aims to let engineering teams focus on agent behavior rather than rebuilding common infrastructure, lowering the production deployment bar for the growing population of teams building agentic applications.