Anthropic launched Claude Managed Agents in public beta on 8 April, offering developers a fully managed cloud service for building and deploying production-grade AI agents without the months of infrastructure work typically required. The service handles the entire agent runtime — spinning up isolated containers, orchestrating tool access, managing state, and recovering from interruptions — so teams can focus on defining what their agents do rather than building the scaffolding to run them.
The technical architecture is designed around practical production concerns. Each agent runs in an automatically provisioned isolated container, with developers pre-specifying the software components needed (browsers for design agents, code interpreters for development agents, and so on). Tool integration is handled by the service itself — developers specify which third-party applications their agents should access, and Managed Agents determines when and how to use each tool during execution. Sensitive data like login credentials for cloud tools is managed through the platform's state management layer.
Pricing follows a straightforward model: standard Claude model usage charges plus $0.08 per agent runtime hour for the managed infrastructure. In research preview, the service also offers multi-agent spawning for complex tasks and automatic prompt refinement, which Anthropic says improved task success rates by up to 10 percentage points in testing. Early enterprise adopters include Notion, Rakuten Group, and Asana, with several having already integrated Managed Agents into their shipping products.
For context engineers, this is a significant shift in the agentic development landscape. Until now, building production AI agents meant assembling your own execution environment — containers, tool routing, error handling, observability, and state persistence — often taking months before the agent itself could be tested at scale. Managed Agents compresses that to weeks. The $0.08 per hour runtime cost also establishes a concrete price point for hosted agent infrastructure, which will shape how teams budget and architect multi-agent systems going forward.
