Anthropic has launched the public beta of Claude Managed Agents, a new platform that allows developers to build and deploy AI agents without having to manually assemble the complex backend infrastructure typically required for agentic systems. The service provides managed execution environments, memory systems, tool permissions, sandboxed runtimes, and operational controls out of the box, enabling developers to focus on defining tasks, tools, and guardrails rather than building orchestration layers themselves. Anthropic says agents can run autonomously in the cloud for extended periods while maintaining state and security controls. The platform also includes early support for coordinated multi-agent workflows, allowing one agent to delegate subtasks to others. Early adopters cited by Anthropic include Notion, Rakuten, Asana, and Sentry, with Rakuten reportedly deploying internal agents across multiple departments in roughly a week each. Anthropic is pricing the service on a usage basis, including an infrastructure fee of approximately $0.08 per agent-hour in addition to standard model consumption charges. Why it matters: Anthropic is moving further up the enterprise AI stack—from selling models to selling managed agent infrastructure. By abstracting away the operational complexity of agent deployment, the company is positioning Claude not just as an AI model provider, but as a full platform for production-grade autonomous AI systems.