Anthropic has launched Managed Agents on its Claude platform, a managed execution layer that separates agent logic from runtime infrastructure concerns. Developers define agent behavior, tools, and constraints while the platform handles orchestration, sandboxing, session state management, credential handling, and persistence. The system uses a meta-harness architecture supporting long-running multi-step workflows with error recovery and session continuity. Priced at $0.08 per session hour, it aims to reduce time-to-production from months to days. Community reactions are mixed — some praise the reduced infrastructure burden, while others raise concerns about vendor lock-in and lack of open standards for portability.
Sort: