LangChain
@LangChain
Managed Deep Agents: Managed, model-agnostic infra for deep agents you can deploy with a single line of code. https://t.co/wiQVO5luru
1retweets23likes
View on XLangChain Launches Managed Deep Agents for Production-Ready Agent Deployment
LangChain· Updated
LangChain introduced Managed Deep Agents in private beta, offering a hosted runtime for deploying deep agents with durable execution and integrated observability. This aims to simplify the operational challenges of running autonomous AI agents in production, allowing developers to focus on agent behavior rather than infrastructure.
LangChain introduced Managed Deep Agents in private beta, an API-first hosted runtime designed for creating, running, and operating deep agents. This offering provides managed, model-agnostic infrastructure for agents built using LangChain's open-source Deep Agents harness, enabling deployment with a single line of code. It includes durable threads, checkpointing, and human-in-the-loop (HITL) workflows.
- Availability
- Private beta
- API Endpoint
- /v1/deepagents
- Supported Workflows
- Durable threads, streaming runs, checkpointing, human-in-the-loop
- Context Management
- Context Hub for persistent agent context and files
- Observability
- Automatic tracing in LangSmith
- Execution Environment
- Sandbox-backed execution for code, shell commands, file I/O
Running long-horizon agents in production requires significant infrastructure for durable execution, memory, tool access, and sandboxes, which often diverts development resources from building the agent itself. Managed Deep Agents addresses this by packaging the operational layer, letting developers focus on agent behavior. Other companies offer managed agent services too, like AWS, Google, and Anthropic.
Managed Deep Agents allows you to create and operate agents programmatically via an API, with agent definitions (instructions, skills, subagents, tool configuration) stored and versioned in LangSmith. The Context Hub provides a managed place for agents to retain and update context across runs, enabling them to improve from real usage. All runs are automatically traced in LangSmith for debugging and performance review.
Still wondering? A few quick answers below.
Managed Deep Agents is an API-first hosted runtime from LangChain that provides the infrastructure needed to run deep agents in production. It handles durable execution, streaming, memory, tool access, and observability, allowing developers to focus on defining the agent's behavior.
It simplifies deployment by managing the underlying runtime infrastructure, including threads, checkpointing, and sandboxes. Developers can define their agent using the open-source Deep Agents harness and then use the Managed Deep Agents API to create, update, and run these agents without building custom servers.
The Context Hub is a managed storage solution within Managed Deep Agents that allows agents to retain and update information across multiple runs. This is crucial for agents that need to learn from interactions, remember user preferences, or maintain project details over time.
Managed Deep Agents is designed for agents that need to work over long time horizons, use tools, preserve context, and produce artifacts. Examples include support and triage agents, research agents, coding agents, data analysis agents, and internal operations agents.
Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →
