Anthropic has come out with a new feature called Claude Managed Agents, and it’s clearly aimed at developers who are trying to turn AI ideas into something that actually works in the real world.
A lot of teams today can build basic AI agents, but getting them ready for real use is where things usually slow down. There’s infrastructure to manage, permissions to set, and a lot of moving parts that aren’t always easy to handle.
Managed Agents tries to ease that burden. Instead of building everything from scratch, developers can use Anthropic’s setup to run their agents. The feature is available through the Claude platform, its console, and even a command-line interface, so it fits into different workflows.
The pricing is fairly straightforward. It follows the usual token-based model, with an extra charge of $0.08 per session-hour while the agent is running.
What stands out here is the idea of letting developers focus on the “what” instead of the “how.” They can define what the agent is supposed to do, what tools it can use, and what limits it should stay within. The rest—like keeping the system stable or managing sessions—is handled in the background.
The platform also supports long-running tasks, meaning agents don’t have to stop and wait for input all the time. They can keep working on their own for longer stretches. There are also built-in features like secure execution, authentication, and tool integration.
Anthropic is also experimenting with a setup where multiple agents can work together. In this case, one agent can hand off tasks to another. It’s still in an early stage, but it gives a sense of where things might be heading.
To keep everything under control, the system includes monitoring and governance tools. Developers can track what their agents are doing, manage permissions, and review activity through the Claude console.
The company says its models are tuned for this kind of workflow. In some cases, agents can even check their own progress and adjust their approach until they reach the desired outcome. That part, however, is still being tested.
In its internal tests, Anthropic noticed better results in structured tasks like file generation compared to more traditional prompting.
Some companies have already started using the feature. Notion is working on bringing it into its platform so users can hand off tasks to AI agents. Rakuten, on the other hand, is using it across different parts of its business, connecting it with tools like Slack and Microsoft Teams.
Also Read: Anthropic Accidentally Exposes Claude Code Source Code
