Agentic Coding with Claude Code
Eden Marco
851人在讀
35.6萬字總字數
¥91.55定價
Eden Marco
Most developers encounter Claude Code through chat-style prompting, but that approach breaks down as projects grow and automation must be safe, repeatable, and controlled. Agentic Coding with Claude Code shows how to move beyond ad hoc prompts and use Claude Code as an extensible, agent-driven development platform. This book focuses on building context-aware AI workflows directly in your terminal and IDE. You will learn how to control Claude Code using slash commands, manage long-term context with persistent memory files, and automate development tasks using hooks that trigger actions across Claude Code's lifecycle. The book also covers the Model Context Protocol (MCP) as an important part of the modern agentic ecosystem. You will understand why MCP exists, explore its core architecture, and configure MCP servers inside Claude Code to improve context sharing across tools, agents, and workflows. The trade-offs between MCP, skills, and subagents are discussed to help you choose the right approach. You will design and orchestrate multi-agent systems using subagents, parallel sessions, and hierarchical delegation. By the end of the book, you will be able to integrate Claude Code into real-world development workflows with confidence and control.
這本書的概念講得不錯，尤其像 skills、MCP、subagent（隔離性、並行性、職能單一性等方面）。而且 deepagent 大概是最後講的，比較新，包括了 plan、subagent, file system access, system prompt 的組合。對 agent技術的各方面都有不少的啓發。
I. The book is organized into three parts. Part I introduces the concept of Context Engineering and the foundations of Claude Code. Part II covers the core components of Claude Code, including MCP, plugins, subagents, and output formats. Part III discusses advanced techniques such as skills, parallel agents, and deep agents. II. The structure is somewhat confusing. Skills, in particular, clearly belong in Part II rather than Part III, and the GitLab workflow would fit better in Part III. Also, the book is based on Claude Code 2.1.77, while the current release is 2.1.227. Since many new features have been added since then, the book largely documents an earlier stage of Claude Code. III. The book offers developers a useful guide to moving beyond chat-style prompts and building safe, repeatable, and scalable automation workflows on top of Claude Code as an agent-driven development platform. However, it is not yet mature enough to be applied directly in production environments.
Just broadly explaining main ideas/concepts around LLM, not fully focusing on coding part.