Anthropic Introduces Three New Features for Claude Managed Agents

AI company Anthropic launched Claude Managed Agents last month, aiming to dramatically simplify the building and deployment of cloud-based AI agents. This week, the platform received three significant updates: Dreaming, Outcomes, and Multiagent Orchestration, further boosting agent intelligence and task-handling capabilities.

Dreaming is a scheduled process that reviews an agent’s historical sessions and memory storage, extracts behavioral patterns, and refines memory content—enabling the agent to self-improve over time. According to Anthropic, Dreaming works in concert with the memory mechanism to significantly extend Claude’s long-term recall abilities. Outcomes allows users to clearly define what constitutes a “successful result,” so agents can more accurately judge whether a goal has been achieved during task execution, improving task completion quality and controllability.

Multiagent Orchestration is the most notable update in this release. It enables a master agent to break complex tasks into subtasks and delegate them to specialized expert agents—each with its own model, prompt, and tools—for collaborative processing. This multi-agent orchestration mechanism allows Claude Managed Agents to handle larger-scale, more intricate business scenarios. As enterprise demand for AI applications continues to grow, Anthropic’s latest updates further strengthen its competitive position in cloud-hosted intelligent agents, with potential to deliver greater value in areas such as automated operations, customer service, and data analysis.