Anthropic updates Claude Managed Agents with three new features

Anthropic updated Claude Managed Agents with three new capabilities: dreaming, outcomes, and multiagent orchestration.

Dreaming is a scheduled memory-refinement preview meant to help agents improve recall over time.

Outcomes adds rubric-based evaluation with separate grading so success criteria are clearer and easier to track.

Multiagent orchestration lets a lead agent coordinate specialist subagents in parallel, making complex workflows more scalable.