A field guide · Claude Managed Agents

Deploy AI workers.
Own the outcome.

The field guide to Anthropic Claude Managed Agents. Fourteen chapters, six diagrams, seven templates — from first-agent mental model to production-ready patterns, grounded in the official docs.

Multi-Agent Delegation — Coordinator and Specialist Sub-Agents A coordinator agent at the top branches to three specialist sub-agents: Researcher, Writer, and QA. Task packets flow down from coordinator to specialists, results flow back up. A shared filesystem is visualized as a floor all agents stand on. Shows the coordinator plus specialists pattern in Claude Managed Agents. Multi-Agent Delegation Coordinator spawns specialist sub-agents via callable_agents — brains can pass hands to one another Shared Filesystem — /workspace all agents read and write via bash · read · write · glob · grep C Coordinator claude-opus-4-7 orchestrates sub-agents R Researcher web_search web_fetch session.thread_created W Writer bash · read write · edit session.thread_idle QA QA Reviewer read · grep bash tests session.thread_idle research task agent.thread_message_sent ↓ ↑ agent.thread_message_received
V2 Preview · Multi-agent delegation
How this book is structured

Four parts. Fourteen chapters. One mental model.

Start at the top and go straight through, or jump to the part you need. Each chapter ends with a hands-on walkthrough and a downloadable template.

PART 01

Foundations

Build the mental model — what agents are, the four building blocks, brain vs. hands, and the three Claude models.

PART 02

Build Your First Agent

Get to a working agent: prototype in Console, define your agent, use the 8 built-in tools, set permission policies.

PART 03

Run Your Agent

Sessions, the event stream, and steering live work — interrupting, redirecting, and approving.

PART 04

Give Your Agent Memory

Why memory matters, the six memory tools, and the patterns that actually work.