Scheduled Agents, Dreaming and Q&A: Claude Managed Agents (Part 3 of 4) — Anthropic

Gabe Cemaj and Priyanka Phatak present a technical demonstration of a platform designed for orchestrating multi-agent AI workflows. The system utilizes an API-first architecture, enabling developers to build custom UIs and schedule recurring tasks through deployments. A core innovation is the "dreaming" feature, which allows agents to autonomously review past session transcripts to refine memory stores and improve future performance. The discussion addresses technical questions regarding observability, multi-tenancy, and scalability, emphasizing the use of workspaces as a primitive for managing complex organizational needs. While the current platform relies on Anthropic models, the speakers highlight its flexibility for building custom applications, noting that enterprise-grade features like role-based access control and potential on-premise deployments are currently on the development roadmap to support broader integration and security requirements.