Claude Managed Agents

What is Claude Managed Agents?

Claude Managed Agents is a service within Claude Platform for running configurable Claude agents on long-running or asynchronous tasks. It supplies the agent execution loop and runtime, coordinating model responses and tool calls.

Key Capabilities

Agent configuration: Developers define a Claude model, system instructions, tools, Model Context Protocol (MCP) servers, and skills. Saved configurations are versioned and reusable across sessions.
Execution environments: Tools can run in Anthropic-managed cloud sandboxes or self-hosted sandboxes. With self-hosting, Anthropic retains model execution and orchestration; customer-operated workers execute tools and return results.
Tools and integrations: Agents can run shell commands, read and edit files, retrieve web content, and use tools exposed by connected MCP servers.
Stateful sessions: Sessions retain conversation history and execution state across interactions. Applications can send further instructions, interrupt work, and retrieve persisted events.
Coordination and oversight: Agents can delegate work to separately configured agents within a session. Event streams and Console views expose execution activity, while configurable permission policies can require approval before supported tools run.

Audience & Use Cases

Audience: Developers and platform engineers building applications that delegate work to agents.
Use cases: Multi-step research, file processing, coding tasks, and workflows requiring repeated tool calls or coordination between specialized agents.

Technical Specifications