Building Production AI Agents for the Web: Design and orchestrate autonomous agents with MCP, Multi-Agent Patterns, and Harness Engineering
暫譯: 打造適用於生產環境的 Web AI Agent：運用 MCP、多代理模式與 Harness Engineering 設計並編排自主代理層

Architect, orchestrate, and deploy scalable AI agents with real-world patterns, multi-agent systems, and evaluation frameworks.

Key Features:

- Filled with real-world agentic web application examples using LLMs, RAG, and tool calling

- Design and implement AI assistants with reasoning patterns like ReAct

- Architect scalable agent systems with planning, orchestration, and autonomy

Book Description:

Adding an LLM to an application is easy. Building an intelligent application that can reliably use your data, call tools, make decisions, complete tasks, and operate in production is much harder.

Building Production AI Agents for the Web is a practical guide to engineering modern AI-powered applications beyond the basic chatbot. You'll start by integrating LLMs into web applications and progressively extend their capabilities with tool calling, RAG, and MCP. From there, you'll learn how these building blocks come together to create AI assistants and agents capable of reasoning and taking action. You'll explore the ReAct architecture, build your first agentic application, and progress toward advanced agent architectures and multi-agent systems. You'll learn proven agent patterns and anti-patterns that help you make better architectural decisions as your applications scale.

In the later chapters, you'll learn how to test LLMs and agents, manage APIs, deploy agentic applications, apply responsible AI practices, and build an agent harness that provides the structure and guardrails agents need to operate reliably.

By the end, you'll understand the complete journey from integrating your first LLM to engineering production-ready intelligent applications that can retrieve information, use tools, reason, collaborate, and take action.

What You Will Learn:

- Integrate LLMs into web apps and extend their capabilities with tool calling and external systems

- Build RAG pipelines that ground AI applications in relevant business and application data

- Use MCP to connect AI applications with tools, resources, and capabilities

- Build intelligent AI agents using ReAct, advanced agent architectures, and multi-agent patterns

- Test and evaluate LLMs and agents to improve reliability and accuracy

- Build an agent harness with the context, tools, guardrails, and infrastructure needed to take AI applications into production

Who this book is for:

This book is for full-stack developers, software architects, and AI practitioners looking to build intelligent, agent-driven web applications. It's also valuable for tech leads and product managers working on AI-powered products.

If you have a basic understanding of web development and want to move beyond simple AI integrations into autonomous, production-ready agent systems, this book will give you the frameworks, tools, and hands-on experience to get there.

Table of Contents

- Introduction to Agentic Apps

- Building Responsible AI Systems

- Introduction to Large Language Models (LLMs)

- Building with LLMs

- Enhancing LLMs with Tool Calling

- Introduction to Retrieval-Augmented Generation (RAG)

- GraphRAG, a Retrieval Approach for Relationship-Heavy Questions

- The ReAct Pattern

- Building an AI Assistant

- Agent Architecture: From AI Assistants to Agents

- From Agent Architecture to Planning Systems

- Agent Orchestration and Autonomy

- Multi-Agent Collaboration and Communication

- Multi-Agent Patterns

- Agent Design Patterns and Anti-Patterns

- Testing LLMs and Agents with Evaluation Frameworks

- Building an Agent Harness

- Preparing Agents for Production

- Operating and Improving Agents in Production