LangChain launched Managed Deep Agents, a production agent-building platform that bundles the Deep Agents harness with managed infrastructure. The move addresses a key friction point: developers building agents currently must assemble a harness, business logic, runtime (durable execution/resumable runs), UX layer (event streaming to UI), sandboxes (for untrusted code), context management (AGENTS.md standards, MCP), evaluation, memory, and auth separately.
Managed Deep Agents represents agents as files and supports custom middleware, custom tools, and programmatic API creation. The bundle includes LangSmith Deployments for runtime, LangSmith Agent Server + Channels for UI streaming, and sandbox infrastructure. This shifts the developer burden: instead of choosing a harness *and* wiring all nine operational concerns individually, developers now pick an integrated managed-harness-and-infrastructure bundle.
The landscape has fractured into several approaches: Claude Managed Agents (API-first for developers), Vercel Eve (file-based agent representation, infrastructure bundled), LangChain Fleet (UI-based for non-developers, extreme "managed" posture), and now Managed Deep Agents (file-based, custom tooling for developers). The convergence on standards like AGENTS.md, MCP, and skills suggests the agent-harness market is stratifying by developer persona and operational maturity.