What managed agents are ## What managed agents are In this channel's material, managed agents refers to Anthropic's Claude Managed Agents — a cloud-hosted agent platform where the agent runs with memory, tools, and the surrounding infrastructure handled for you. As the hosts put it in a conversation with the team who built it (Angela Jiang, head of product on the Claude platform, and Katelyn Lesse, head of engineering), the platform evolved from a completion endpoint into exactly this: Claude in a cloud computer with memory, tools, and infrastructure.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@1181] ## Why managed infrastructure matters Why managed infrastructure matters Teams running agents in production expect harness engineering to be the bottleneck, but Katelyn Lesse says the real wall everyone hits is production infrastructure — keeping servers running, storing transcript data, and secure sandboxing, where a dying sandbox takes your whole agent with it.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@1181] Harnesses also performed drastically differently on evals, and new models require harness and model to be very paired rather than hot-swappable.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@813] That is why managed agent platforms may be sneaky important: Anthropic ships a scalable cloud-hosted agent platform with memory and orchestration. Inside a bigger company, a Mac mini going down is unacceptable; reliability, not peak performance, is the real value a managed platform provides.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@692] How managed agents are used ## How managed agents are used Across the channel's material, the direction of travel is from synchronous prompting to asynchronous delegation: you hand a task off, the agent spends significant time making calls, integrating responses, and iterating, then reports back. Claude Managed Agents launched with outcome-based tasks where the agent works asynchronously and returns only when done.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@813] The team's framing is that agents should eventually be judged by just two parameters: an outcome and a budget.[^4872db36-e98a-5f6b-bc5c-b11e48c9411b@692] - Managed agents means Claude running on a cloud-hosted platform with memory, tools, and infrastructure managed for you. - The value is reliability in production — uptime, transcript storage, sandboxing — not peak performance. - Usage is asynchronous and outcome-based: hand off a task with an outcome and a budget; the agent returns when done. - Anthropic's managed platform stands in contrast to an agents SDK without a cloud version, which is why the hosts argue managed agent platforms may be sneaky important.