Claude Managed Agents: Anthropic Redefines AI Development

Anthropic recently unveiled a groundbreaking feature called Claude Managed Agents. This interface allows Claude, the artificial intelligence developed by the company, to create and manage multiple AI sub-agents to autonomously accomplish complex tasks. This technological advancement enables the delegation of an entire project, whether it is computational or analytical, without requiring human intervention between different stages.

With this update, Claude now takes on the role of project manager within the Anthropic platform. Available to Pro and Team subscribers, the Managed Agents feature allows users to submit a global request. Claude then generates autonomous instances that independently address each aspect of the problem. These sub-agents operate in isolated environments, using specific tools such as code execution or web searching, before relaying their results back to the main agent. This architecture eliminates the need for third-party frameworks like LangChain, as orchestration occurs directly within the model.

Claude Managed Agents: A Revolution in Project Management

The introduction of these autonomous agents transforms the role of developers in software production. Traditionally, engineers had to manually build interactions between different AI components, define permissions, and monitor feedback loops. Now, Claude makes these architectural decisions in real-time. Companies like Sentry have already adopted this system, moving from bug reporting to a ready-to-merge correction proposal in a single command.

While this technology promises increased productivity for businesses, it also poses a threat to mid-level technical profiles, whose immediate utility is called into question. Claude proposes code and assembles a virtual team to deploy it, thereby reducing human supervisors' visibility over decision-making stages. Although a logbook is available for each active session, leaders can now delegate the structure of their workflow to an intelligence capable of self-organizing. This transfer of skills to the machine results in a loss of substance for technical management positions and, consequently, a reduction in human resources.

A New AI Economy

Companies capable of automating their structure stand out thanks to access to this paid technology. Anthropic charges for the use of Claude Managed Agents at $0.08, or about €0.07 per hour of active session, in addition to the tokens consumed. Companies like Rakuten and Notion are already using these customized agents to accelerate their internal work cycles by a factor of ten. An employee can simply ask Claude to conduct a comprehensive competitive analysis, and the AI activates five simultaneous agents to analyze global markets.

This system allows users to focus on other tasks while the agents work in parallel. However, it also reinforces dependence on a single provider, with centralized infrastructure, intelligence, and rights management. Early production tests show that these managed agents successfully complete complex tasks with a ten-point higher success rate compared to traditional methods. The AI now awaits instructions from its own digital reflections rather than from humans to progress. However, this proliferation of agents working in a closed ecosystem could saturate enterprise network bandwidth. Although these agents are invisible and can be tasked endlessly, they are not free.