Anthropic has launched Claude Managed Agents in public beta. The API suite is designed to let developers build and run cloud-hosted AI agents without having to create their own infrastructure for sandboxing, state management, or tool execution.

Until now, teams shipping production-grade agents typically had to build secure containers, permission systems, and custom agent loops themselves. Managed Agents is meant to handle that overhead. According to Anthropic’s documentation, the system provides an orchestration harness that can call tools on its own, manage context, and recover from errors. Anthropic says this can reduce the time from prototype to production by as much as 10x.

Sessions can run autonomously for hours, and results persist even if the connection drops. Built-in tools include Bash commands, file operations, web search, and integrations with external services through MCP servers. In a separate Research Preview that is not yet broadly available, agents can also launch other agents, coordinate parallel tasks, evaluate outputs, and manage memory. Developers interested in those features can join a waitlist.

Anthropic says several enterprise customers are already using the platform. Notion allows teams to delegate tasks directly to Claude inside the workspace. Rakuten has built enterprise agents for sales, marketing, and finance, integrated into Slack and Microsoft Teams, and says each was deployed within about a week. Sentry connected its debugging agent to a Claude agent that writes patches and opens pull requests.

Managed Agents is available to all API accounts and requires a specific beta header, which Anthropic’s SDK sets automatically. Anthropic also notes that behavior may change between releases. Pricing is usage-based: standard token charges apply, plus $0.08 per session hour.

For now, the system runs only on Anthropic’s own infrastructure. The announcement does not say whether it will also become available through partner platforms such as Amazon Bedrock or Google Vertex AI. That could be a meaningful limitation for enterprises pursuing multi-cloud strategies