Anthropic debuts Dreams for Claude Managed Agents

Anthropic launched Dreams, a research preview for Claude Managed Agents that analyzes agent memory to spot patterns and correct recurring mistakes.

The update adds rubric-based outcomes tracking and multiagent orchestration to improve how agents are evaluated and coordinated.

Developers can access the new capabilities through the Claude Platform.

The changes are aimed at making AI agents more reliable for complex enterprise workflows.