The company says the deepagents deploy command bundles agent orchestration and memory with a horizontally scalable LangSmith Deployment server. Builders specify a model, AGENTS.md instructions, skills, MCP tools and, optionally, a sandbox. The announcement lists integrations with multiple model and sandbox providers, and endpoints including MCP, A2A, Agent Protocol, human-in-the-loop and memory APIs. LangChain says agents can be self-hosted and that memory is stored in a standard format and can be queried directly. The launch positions open harness and memory control as alternatives to a managed, proprietary agent stack. These are vendor-described capabilities; the announcement does not establish comparative portability, operational performance or the practical effort required to migrate an agent and its accumulated memory.
Deep Agents Deploy: an open alternative to Claude Managed Agents
In one minute
LangChain announced Deep Agents Deploy beta, a deployment service for agents built with its open-source, model-agnostic Deep Agents harness.