Anthropic Launches Managed Agents API in Beta

Anthropic released Claude Managed Agents in public beta - a hosted API layer for building production agents with sandboxed execution, persistent sessions, and multi-agent coordination. The pitch is cutting months of infrastructure work down to days by handling state management, permissioning, and credential routing out of the box. Notion, Rakuten, Asana, and Sentry are already shipping on it, with use cases ranging from Slack-based enterprise agents to automated PR debugging. Internal benchmarks show a 10-point task success improvement over standard prompting loops on complex problems. Multi-agent delegation and self-evaluation are gated behind a research preview. Every major AI lab is now racing to own the agent deployment layer - OpenAI with Codex and Agents SDK, Google with Jules and Vertex - and Anthropic just made its infrastructure play explicit.