AI agents and Harnesses foundations: Building from ReAct Loops to Long Horizon Agent Harnesses with LangChain and LangGraph-电子书百科大全

AI agents and Harnesses foundations: Building from ReAct Loops to Long Horizon Agent Harnesses with LangChain and LangGraph book cover

AI agents and Harnesses foundations: Building from ReAct Loops to Long Horizon Agent Harnesses with LangChain and LangGraph

Author(s): Eden Marco (Author)

Publisher: Packt Publishing - ebooks Account
Publication Date: October 9, 2026
Edition: 1st
Language: English
Print length: 205 pages
ISBN-10: 1808087313
ISBN-13: 9781808087318

Book Description

“Harness design is key to performance at the frontier of agentic coding.” — Anthropic Engineering, 2026

Go beyond simple prompts and build production-grade AI agents with LangChain and LangGraph, from ReAct reasoning loops to tool-calling, RAG, and context engineering.

A hands-on LangChain guide to building agents and the harnesses that run them

Agents and agent harnesses are everywhere now, in product launches, keynotes, and job descriptions. Most developers using them couldn't tell you where the model stops and the harness starts.

That distinction is the whole book. A language model on its own has no memory, no filesystem, no clock, and no ability to act, it is a stateless function that takes messages and returns text or a tool request. Everything that makes it feel like an agent, remembering what it said, reading your files, retrying after failure, knowing when to stop, comes from code written around it. This book builds that code by hand before you ever use a framework's version of it.

In this book, you'll learn how to:

Tell agents and harnesses apart by building the agent loop yourself: call the model, run the tools it requests, feed results back, repeat
Build a working AI agent with tool-calling from a single API call up to a full agent loop
Rebuild agents without LangChain using handwritten JSON schemas, then raw prompts and regex parsing, the way early agents were built before structured tool calling existed
Trace and debug agent behavior end to end using LangSmith instead of guessing why an agent failed
Apply prompt and context engineering to fix the failures that show up once an agent runs for a while
Understand what a harness actually supplies context management, persistence, guardrails, and delegation, and see it in LangChain's Deep Agents

Why this book stands out

Most agent tutorials show you the decorator and stop there. This one shows you what the decorator, and the harness around it, is actually doing.

Built from first principles: every layer of abstraction, from LangChain down to raw prompts and regex, is stripped away and rebuilt by hand
Harnesses explained, not name-dropped: the book defines what a harness supplies and traces it through real systems like Claude Cowork, Codex, and LangChain's Deep Agents
Grounded in where the field came from: LangChain's role in popularizing agent abstractions, and how that lineage runs through LangGraph and Deep Agents today
Theory that actually explains /li>

Who this is for

Developers who've used LangChain, or agent tools like Claude Code or Codex, and want to understand what's happening beneath the vocabulary, harnesses, subagents, context engineering, MCP, skills, before building anything serious with it. Basic Python and familiarity with LLM APIs assumed. This is a foundations book: it builds the grounding, not a production deployment playbook.

By the end, you won't just type a sentence and watch a pull request appear. You'll know what happened between those two moments, and why.

Introduction
The GIST of LangChain- Get started by with your "Hello World" chain
THE GIST Of AI Agents
Agents Under The Hood
[Layer 2] Manual JSON Schemas vs. LangChain Tool Abstraction
[Layer 3] The ReAct Prompt: The Foundation of AI Agents
Function Calling
Prompt Engineering Theory
Deep Agents

Editorial Reviews

About the Author

Eden Marco is an LLM specialist at Google Cloud and a LangChain Ambassador with years of experience in software engineering and cloud architecture. He was one of the first engineers at Orca Security, holds a bachelor's degree in computer science from the Technion, and has taught at Reichman University. Eden creates practical, production-ready courses based on real-world experience.

View on Amazon

{"@context":"https://schema.org","@type":"Book","name":"AI agents and Harnesses foundations: Building from ReAct Loops to Long Horizon Agent Harnesses with LangChain and LangGraph","image":"https://m.media-amazon.com/images/I/41X0mcyLQ+L._SX342_SY445_FMwebp_.jpg","author":{"@type":"Person","name":"Eden Marco (Author)"},"publisher":{"@type":"Organization","name":"Packt Publishing - ebooks Account"},"datePublished":"October 9, 2026","isbn":"9781808087318","numberOfPages":205,"inLanguage":"English","description":"“Harness design is key to performance at the frontier of agentic coding.” — Anthropic Engineering, 2026Go beyond simple prompts and build production-grade AI agents with LangChain and LangGraph, from ReAct reasoning loops to tool-calling, RAG, and context engineering.A hands-on LangChain guide to building agents and the harnesses that run themAgents and agent harnesses are everywhere now, in product launches, keynotes, and job descriptions. Most developers using them couldn't tell you where the model stops and the harness starts.That distinction is the whole book. A language model on its own has no memory, no filesystem, no clock, and no ability to act, it is a stateless function that takes messages and returns text or a tool request. Everything that makes it feel like an agent, remembering what it said, reading your files, retrying after failure, knowing when to stop, comes from code written around it. This book builds that code by hand before you ever use a framework's version of it.In this book, you'll learn how to:Tell agents and harnesses apart by building the agent loop yourself: call the model, run the tools it requests, feed results back, repeatBuild a working AI agent with tool-calling from a single API call up to a full agent loopRebuild agents without LangChain using handwritten JSON schemas, then raw prompts and regex parsing, the way early agents were built before structured tool calling existedTrace and debug agent behavior end to end using LangSmith instead of guessing why an agent failedApply prompt and context engineering to fix the failures that show up once an agent runs for a whileUnderstand what a harness actually supplies context management, persistence, guardrails, and delegation, and see it in LangChain's Deep AgentsWhy this book stands outMost agent tutorials show you the decorator and stop there. This one shows you what the decorator, and the harness around it, is actually doing.Built from first principles: every layer of abstraction, from LangChain down to raw prompts and regex, is stripped away and rebuilt by handHarnesses explained, not name-dropped: the book defines what a harness supplies and traces it through real systems like Claude Cowork, Codex, and LangChain's Deep AgentsGrounded in where the field came from: LangChain's role in popularizing agent abstractions, and how that lineage runs through LangGraph and Deep Agents todayTheory that actually explains /li>Who this is forDevelopers who've used LangChain, or agent tools like Claude Code or Codex, and want to understand what's happening beneath the vocabulary, harnesses, subagents, context engineering, MCP, skills, before building anything serious with it. Basic Python and familiarity with LLM APIs assumed. This is a foundations book: it builds the grounding, not a production deployment playbook.By the end, you won't just type a sentence and watch a pull request appear. You'll know what happened between those two moments, and why.Table of ContentsIntroductionThe GIST of LangChain- Get started by with your \"Hello World\" chainTHE GIST Of AI AgentsAgents Under The Hood[Layer 2] Manual JSON Schemas vs. LangChain Tool Abstraction[Layer 3] The ReAct Prompt: The Foundation of AI AgentsFunction CallingPrompt Engineering TheoryDeep Agents","bookEdition":"1st","url":"https://www.amazon.com/dp/1808087313/","bookFormat":"http://schema.org/EBook","additionalType":"http://schema.org/PDF","fileSize":"13 MB","accessibilityFeature":["login required","member access only"],"accessibilitySummary":"PDF version available to authenticated members only. File size: 13 MB."}