Anthropic16:46Feature UpdatesOfficial X
Claude Managed Agents Add Dreaming and Multiagent Features
Self-improving agents handle complex tasks faster and higher quality.
Key Points
- 1Dreaming curates memories from past sessions
- 2Outcomes self-correct via rubrics
- 3Lead agent delegates to parallel specialists
- 4Webhooks for notifications
Anthropic enhanced Claude Managed Agents with dreaming (research preview), outcomes, multiagent orchestration, and webhooks (public beta). Agents learn from past sessions and handle complex parallel tasks. Developers can build reliable automations easily.