With Claude Managed Agents, Anthropic wants to run your AI agents for you

Anthropic launched Claude Managed Agents in public beta for building and running enterprise AI agents on its hosted infrastructure.

Companies can define agents in natural language or YAML, add guardrails, and connect tools, auth, sandboxing, and third-party services.

Advanced features like memory, multi-agent orchestration, and self-evaluation remain in research preview.

The move pushes Anthropic deeper into the agent infrastructure stack and lowers the effort needed to deploy governed production agents.