Anthropic has introduced new features for its Claude agents, including a 'dreaming' tool aimed at improving AI memory and performance. Dreaming helps agents remember past work, identify recurring mistakes, and refine workflows by extracting patterns and curating memories. This feature is particularly beneficial for long-term projects and multiagent orchestration, allowing agents to retain context across sessions. Additionally, Anthropic has launched outcomes, multiagent orchestration, and webhooks for developers using Managed Agents. Outcomes enable developers to set success criteria, with a grader evaluating and prompting revisions to improve task success rates. Internal tests showed a 10% improvement in task success and enhanced file generation quality. Multiagent orchestration allows a lead agent to delegate tasks to specialist agents, facilitating parallel work on complex projects. These updates aim to enhance the utility of Claude agents in enterprise environments where accuracy and coordination are crucial.