Anthropic Adds Dreaming, Outcomes, and Multiagent Orchestration to Claude Managed Agents

Summary

Anthropic updated Claude Managed Agents with three new capabilities that materially expand what production agentic systems can do. The first, Dreaming (research preview), enables agents to review past sessions and memory stores, extract patterns, and self-improve over time — users choose between automatic memory updates or manual review before changes apply. The second, Outcomes, lets developers define success criteria through rubrics: a separate grader evaluates agent output against specified c