Managed Agents: what they are and what they mean for your organisation
Anthropic just launched Managed Agents. Here is what they do, who they are for, and how to think about whether your team should use them.
Anthropic launched Managed Agents in April 2026. If you have been following the AI space, you have heard the word "agent" a lot. Most of the time it is vague. This is not.
Managed Agents are Claude running in the cloud, completing multi-step tasks on its own, with Anthropic handling the infrastructure. You define what the agent should do. Anthropic runs it, sandboxes it, and gives you the result.
What this actually means in practice
Think of it this way. Right now, when you use Claude, you are in a conversation. You ask, Claude answers. If you want Claude to do something complex — research a market, process a batch of documents, prepare a report that requires multiple steps — you either do it as a long conversation or you build custom infrastructure.
Managed Agents remove the second option's complexity. You describe the task, give the agent the tools it needs (web search, file access, code execution), and it runs. It can take minutes or hours. You get the output when it is done.
The key difference from a regular conversation: the agent works autonomously through multiple steps. It does not wait for you between each one. It decides what to do next based on what it found in the previous step.
Who this is for
Teams building internal tools. If you have developers who want to build AI-powered workflows — processing incoming documents, monitoring data sources, generating reports — Managed Agents handle the infrastructure so your team focuses on the logic.
Operations teams with high-volume research tasks. Competitive analysis, vendor evaluation, market monitoring. Tasks where a human currently spends hours gathering and synthesising information from multiple sources.
Companies that tried building agents and hit infrastructure problems. Sandboxing, error handling, scaling, session management — Managed Agents handle all of this. If you built an agent prototype that worked but was painful to run in production, this is the managed version.
Who this is not for (yet)
Non-technical teams who just use Claude.ai. Managed Agents are currently an API product. You need a developer to set them up. If your team is using Claude through the chat interface, the features that matter to you are Cowork and Dispatch instead.
Tasks that need real-time human judgment. Managed Agents run autonomously. If the task requires a human to review intermediate steps — customer communications, legal documents, anything where a wrong output has real consequences — you need a human-in-the-loop architecture, not a fully autonomous agent.
The cost model
$0.08 per session-hour plus your normal token costs. A session that runs for 30 minutes costs $0.04 plus tokens. For most use cases, the session cost is small compared to the token cost of the work the agent is actually doing.
Compare this to the cost of building and hosting your own agent infrastructure: servers, sandboxing, error recovery, scaling. For most teams, managed is cheaper until you are running thousands of sessions per day.
How to think about it
Managed Agents are not a replacement for Claude conversations. They are a different tool for a different kind of work. Conversations are for interactive, back-and-forth collaboration. Agents are for "go do this thing and come back with the result."
If you are an operator trying to decide whether to explore this: start by identifying the tasks your team does that take more than 30 minutes, involve gathering information from multiple sources, and produce a document or summary at the end. Those are your agent candidates.