Anthropic's Managed Agents — infrastructure for production AI workers
Anthropic published a tutorial building an incident-response agent using Agents, Environments, and Sessions — its new platform for stateful, long-running tasks.
The real shift: agents in production need state persistence, audit logs, permission boundaries, tool sandboxing, memory management, and secure execution before enterprises will deploy them at scale.