LangChain Launches Managed Deep Agents 0.8
At the Interrupt conference, LangChain announced Managed Deep Agents 0.8, aimed at solving four major pain points for production agents: memory, authentication, channels, and tool management—reducing building a production-grade deep agent to "writing a project directory."
Confirmed
- User-held credentials and per-user persistent memory: agents can act with the correct permissions and work from personalized user context, with access-policy controls on memory
- A new HTTP channel that supports webhooks
- Sandboxed file API and human-in-the-loop review capabilities
- Several other announcements accompanied the release: LangSmith Engine v2 (red-teaming and automated testing) and the launch of Deep Agents fine-tuning (summarized from Harrison Chase's keynote)
Why it matters
- Auth isolation, stateful memory, and external channel integration have long been major blockers for shipping agents to production; 0.8 platformizes these capabilities, lowering the barrier to enterprise-grade deployment
2026-09-24 ~ 2026-09-25 · 5 related posts
Primary sources
- LangChain ships Managed Deep Agents 0.8 with user-scoped auth, memory and HTTP channels — LangChain ·
- LangChain's Interrupt keynote ships LangSmith Engine v2, Deep Agents 0.8, fine-tuning and more — LangChain ·
- LangChain ships managed-deepagents 0.8 with agent memory, sandbox, and webhooks — Meta_Engineers ·
- [source] LangChain's Interrupt keynote ships LangSmith Engine v2, Deep Agents 0.8, fine-tuning and more — LangChain · 2026-09-24
- [source] LangChain ships Managed Deep Agents 0.8 with user-scoped auth, memory and HTTP channels — LangChain · 2026-09-25
- LangChain ships Managed Deep Agents 0.8 at Interrupt with sandbox file API and HTTP channels — LangChain · 2026-09-25
- [source] LangChain ships managed-deepagents 0.8 with agent memory, sandbox, and webhooks — Meta_Engineers · 2026-09-25
1 near-duplicate retellings: LangChain