Agentic Engineering Fieldbook

Principles, patterns, and practices for AI-native systems.

This fieldbook is a living knowledge system for engineers who design, build, and operate AI-native software.

It organizes knowledge around enduring engineering questions—architecture, context, evaluation, reliability, product design, and trade-offs—rather than around individual vendors or frameworks.

What we mean by agentic engineering

In this fieldbook, agentic engineering means the practice of designing, building, evaluating, and operating AI-native software systems that exhibit goal-directed behavior, tool use, state management, feedback loops, and bounded autonomy. It is not limited to building a single agent, and it is not the same as using agents to assist software development.

Why a fieldbook?

A fieldbook is a working reference made for use in the field. It collects durable principles, tested patterns, trade-offs, and observations that can be consulted during real work and revised as practice changes. The name signals that this project is neither a textbook claiming a final theory nor a vendor manual tied to one tool.

The Chinese title, 《智能体工程实践手册》, is a functional rather than literal translation. “工程” preserves the software-engineering perspective, while “实践” carries the sense of knowledge formed and used in real settings. In Chinese chapters, we use “本手册” for ordinary references and retain Fieldbook when discussing the project’s name or editorial philosophy.

Current stage

Version 0.1 establishes the content framework, editorial method, evidence policy, translation workflow, and the first complete handbook structure: Foundations, Architecture, Production Engineering, Product & UX, Patterns & Cases, and maintained Field Notes.