LangChain said on Sept. 24 that Managed Deep Agents 0.8 adds user-owned credentials, user-level memory, HTTP channels, Slack file transfer and built-in web search.
The update targets production concerns that sit outside prompt design, including identity scoping, channel access and separation of shared versus personal context, which can determine whether an agent system fits internal tools or customer-facing workflows.
According to the company's blog post, Managed Deep Agents packages LangChain's Deep Agents harness with managed infrastructure for deployment, including channels, authentication and tool permissions. LangChain said the release addresses gaps between prototyping with the harness and operating agents in production, where auth, memory and channel plumbing can require substantial engineering work; Automatica Press has separately reported on how operational burden, not benchmark scores alone, can determine whether agent products hold up in practice.
LangChain said the 0.8 release introduces a second memory layer keyed to an authenticated user alongside existing agent-level durable memory. In the company's description, agent memory is mounted at /memories/agent/ and shared across a deployment, while user memory is mounted at /memories/user/ and tied to the caller's identity, with the runtime not copying content between the two layers, according to the LangChain blog post.
The company also said Managed Deep Agents now supports user-owned and agent-owned credentials through "connections" to external services such as GitHub, Notion and Tavily. LangChain said the service includes out-of-the-box support for 23 services including Linear, GitHub and Google Workspace tools, with LangSmith managing authorization, tokens and methods, according to the blog post.
For access paths, LangChain said the release adds HTTP channels for any service that can send a JSON webhook and expands Slack support with file transfer. The post says Slack users can invoke an agent from a direct message, app mention or thread reply, while HTTP channels are intended for internal tools, customer portals, support systems and other webhook-capable product surfaces, according to the blog post.
LangChain further said Managed Deep Agents now includes web search powered by Parallel without requiring a separate vendor account or API key, and that the tool is available free while Managed Deep Agents remains in beta, according to the blog post. These functionality claims come from LangChain's own announcement, which includes no independent test results.