Book
Tsukutte Manabu AI Agent (Learning AI Agents by Building One)
Opening Up AI-Era Engineering with TypeScript and LLMs
laiso
Summary
A Japanese-language practical guide by laiso to building AI coding agents, not just using them. Agent-driven coding, where an agent reads files, runs commands and tests, checks results, and repeats fixes as needed, is becoming mainstream in software development. Working mainly in TypeScript and Bun, you build a coding agent that starts from a GitHub issue and automates fixing code through opening a pull request, including GitHub integration, while implementing its own reasoning mechanism and learning to predict, control, and customize its behavior. The agent you build, called "Nano Code" by the author, is assembled chapter by chapter: an LLM API connection, tools for files and commands, a reasoning loop, Git operations, and integration into a working environment. It targets engineers who have built TypeScript applications or CLI tools and are interested in AI agents and LLMs.
Target Readers
- Web engineers who have used TypeScript
- Engineers interested in AI agent development and using LLMs
- Engineers who want to learn new technologies and development methods
Tags
Colophon
- Publisher
- Gijutsu-Hyohron
- ISBN
- 978-4-297-15565-0
- Published
- Apr 2026
- List price
- ¥3,520incl. taxMay differ from the actual selling price on Amazon
Get this book
* The link above is an advertisement via Amazon Associates.Related Books
Prerequisites
- Recommended
Context Engineering
Understood through LLM Principles, RAG, and Agent Development
Hirosato Gamo
Reason: Understanding how to call LLM APIs and the basics of context design lets you keep sight of what's happening behind the reasoning loop and tool calls when building an agent from scratch in TypeScript. The context-engineering knowledge underpins the design decisions behind a self-built coding agent like Nano Code.
- Recommended
Kaihatsu Koritsu o Up Suru! Claude Code Jitsuyo Nyumon (A Practical Introduction to Claude Code for Boosting Development Efficiency)
Fumitaka Osawa
Reason: Once you can use an AI coding agent like Claude Code effectively, you start wanting to build one yourself to see what's happening inside. Building AI Agents walks through implementing everything from LLM calls to tool execution and Git/GitHub integration, shifting your perspective from using such a tool to building one.
- Recommended
Jissen Claude Code Nyumon (A Practical Introduction to Claude Code)
A Way of Thinking about AI Coding for Use in Real-World Development
Masahiro Nishimi, Shingo Yoshida, and Yuki Oshima
Reason: Once you've reframed Claude Code as an 'agent that executes agreements' and internalized the spec-driven development pattern, you naturally want to see with your own hands what happens inside such an agent. Building AI Agents pushes you from being a user to a builder by having you implement the reasoning loop and tool calls in TypeScript yourself.
Next Books
- Recommended
Building Applications with AI Agents
Designing and Implementing Multiagent Systems
Michael Albada
Reason: Having built a coding agent from scratch in TypeScript, you internalize the inner workings of the reasoning loop and tool integration. Practical AI Agent Development builds on that implementation sense, broadening the view from a single agent to multi-agent systems and production operation.