Managed Deep Agents delivers a better user experience for agents in production
LangChain announced the launch of Managed Deep Agents 0.8. According to the company, the release adds support for user-owned credentials, user-level memory, HTTP channels, file transfer in Slack, and a pre-built web search tool powered by Parallel.
Managed Deep Agents combines the Deep Agents harness with infrastructure for running agents in production. A Managed Deep Agent is described as a code-first project in a repository, organizing primitives such as an agent file, instructions, identity, memory, tools, channels, middleware, schedules, connectors, skills, sandbox configuration, and evals into a directory structure.
On memory, LangChain said the release adds a user-level layer scoped to the authenticated person starting an agent run, in addition to existing durable agent memory. Agent memory is mounted at /memories/agent/ and shared across the deployment, while user memory is mounted at /memories/user/ and keyed to the caller's authenticated identity. Memory is backed by LangSmith Context Hub. Defaults allow both layers in Slack one-to-one DMs, while Slack channels or group DMs and HTTP deny user memory by default unless specified otherwise.
Connections link an agent to external services, with credentials scoped at either the user or agent level. LangChain said Managed Deep Agents includes out-of-the-box support for 23 services including Linear, GitHub, and Google Workspace tools.
The release also adds HTTP channels for connecting agents to services that can send a JSON webhook, and file transfer in Slack. For Slack workflows, users can invoke an agent from a DM, an app mention, or a thread reply, and can send files such as logs, spreadsheets, contracts, screenshots, or customer documents.
Web search powered by Parallel is now built into Managed Deep Agents. LangChain said users specify the MCP server in the tools folder and LangSmith manages the Parallel credentials and executes the tool, with calls, latency, and errors appearing in LangSmith traces. LangChain said it is making Parallel web search available free through Managed Deep Agents while MDA is in beta.
The post included comments from Zahid, CTO of Kyth.ai, and Derek Gilbert of Engineering at Consensus, describing their use of the product.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Managed Deep Agents is the simplest way to build, deploy, and run agents in production. The 0.8 release adds support for user-owned credentials, user-level memory, HTTP channels, file transfer in Slack and a pre-built tool for web search.