book

AI Agents in Action, Second Edition

Name: AI Agents in Action, Second Edition
Author: Micheal Lanham
ISBN: 9781633434530

by Micheal Lanham

June 2026

Intermediate

392 pages

11h 35m

English

Manning Publications

Read now

Unlock full access

AI Agents in Action, Second Edition
copyright
contents
preface
acknowledgments
about this book
about the author
about the cover illustration
1 The rise of AI agents
1.1 Defining agents and agentic thinking1.1.1 Understanding agent, assistant, and LLM patterns1.1.2 Thinking like agents: Sense-plan-act-learn1.1.3 Agents act with tools1.2 Introducing the Model Context Protocol1.3 Understanding the five functional layers of an agent1.3.1 The agent persona1.3.2 Agent tools and actions1.3.3 Agent reasoning and planning1.3.4 Agent knowledge and memory1.3.5 Agent evaluation and feedback1.4 Advancing to multi-agent systems1.4.1 The agent flow assembly line1.4.2 Agent orchestrations (hub-and-spoke)1.4.3 Agent collaboration (teams of agents)1.5 Next steps
2 Core components: Large language models, prompting, and agents
2.1 Understanding large language models2.1.1 LLMs: Probabilistic token machines2.1.2 What is a token?2.1.3 Tuning temperature, top-p, and more2.2 Controlling LLMs with prompt engineering (agent persona)2.2.1 Applying core prompt techniques2.2.2 Thinking like an LLM2.2.3 Avoiding common prompt pitfalls2.3 Building an agent with OpenAI Agents2.3.1 Building a minimal agent2.3.2 Setting the agent model and other parameters2.3.3 Controlling inputs and typed outputs2.3.4 Tracing agents2.4 Enhancing agents through tool integration2.4.1 Providing agents with tools2.4.2 Tracing agentic tool use2.5 Exercises

3 Actions with Model Context Protocol for AI agents
3.1 Understanding MCP fundamentals for agent development3.1.1 The standardization problem MCP solves3.1.2 MCP architecture: Clients, servers, and services3.1.3 Core components: Tools, resources, and prompts3.1.4 MCP deployment patterns for agents3.1.5 MCP powers the functional agent layers3.2 Getting started with MCP servers3.2.1 Coding up an MCP server for Claude3.2.2 Using the MCP inspector3.2.3 Understanding MCP transport types3.2.4 From desktop to agents: The key differences3.3 Using MCP servers for agents3.3.1 Using agents with local MCP servers over STDIO3.3.2 Using local MCP servers over SSE with agents3.3.3 Connecting to the standard MCP servers3.4 Building MCP servers for agents3.4.1 Converting tools to an MCP server3.4.2 Consuming MCP servers locally or remotely3.5 Exercises
4 Architecting and building multi-agent systems
4.1 Architecting multi-agent systems4.1.1 Decision-making and control patterns4.1.2 Communicating with shared memory, message passing, and MCP4.1.3 Channeling multi-agent coordination strategies4.2 Balancing agents with agentic flows4.2.1 Transforming agents to agent flows4.2.2 Building an agent-to-agent flow4.2.3 Agency and decision-making in agent flows4.3 Understanding handoffs in agent flows4.3.1 Agent-to-agent flow with handoffs4.3.2 Visualizing agent flows4.3.3 Monitoring the handoff4.4 Validating agent flows with guardrails4.4.1 Implementing input and output guardrails4.4.2 Using agents as guardrails4.4.3 Adding guardrails for pass-off agent flows4.5 Exercises
5 Agent reasoning and planning
5.1 Understanding LLM reasoning and planning5.1.1 Chain-of-thought reasoning5.1.2 Reasoning, acting, observing: The ReAct paradigm5.1.3 Planning with LLMs5.2 Instructing agents to reason and plan5.2.1 Applying CoT to an agent5.2.2 Implementing ReAct with agents5.3 Advanced reasoning patterns with agents5.3.1 Tree-of-thought5.3.2 Reflexion5.3.3 Selecting the right pattern for your agents5.4 Utilizing the sequential thinking MCP server5.4.1 Unchaining the sequential thinking server5.4.2 Revisiting time travel problems with sequential thinking5.4.3 Advanced reasoning with sequential thinking5.5 Exercises
6 Working with memory and knowledge RAG for agents
6.1 Understanding retrieval in AI applications6.1.1 The basics of RAG6.1.2 Delving into semantic search and document indexing6.1.3 Applying vector similarity search6.2 Vector databases and similarity search6.2.1 Demystifying document embeddings6.2.2 Querying document embeddings from Chroma DB6.3 Building practical RAG knowledge agents6.3.1 Everything begins with search and relevance6.3.2 Building a vector search RAG agent6.3.3 Building a hybrid search RAG agent6.4 Adding memory to agents with MCP6.4.1 Understanding memory form and agent function6.4.2 Attaching a graph database for memory using MCP6.4.3 Creating hybrid memory systems with MCP6.4.4 Semantic augmented memory and applications to semantic, episodic, and procedural memory6.4.5 Uncluttering memory with compression and forgetting6.5 Exercises
7 Building robust agents with evaluation and feedback
7.1 Introducing agent evaluation and feedback7.2 Implementing test-driven agent development7.2.1 Exploring TDAD in practice7.2.2 Coding and testing the RAG agent7.2.3 Refactoring the agent7.2.4 Extending evaluation with an agent evaluator7.3 Employing grounding, critic, and evaluation agents7.3.1 Reviewing the grounding agent7.3.2 Grounding the RAG agent7.3.3 Implementing grounding agents as guardrails7.3.4 Understanding the role of rubrics in evaluation7.3.5 Building a rubric critic agent7.4 Phoenix for evaluation and feedback7.4.1 Connecting to Phoenix7.4.2 Adding metadata and session tracking7.4.3 Experimenting with evaluators7.4.4 Providing feedback with annotations7.5 Exercises
8 Deploying agents and agentic systems
8.1 Strategies for consuming agents8.1.1 Embedding real-time voice agents into web applications8.1.2 Hosting agents through an API8.1.3 Consuming an agent web service in a web application8.2 Dockerizing agent systems8.2.1 Containerizing an agent microservice8.2.2 Orchestrating agentic systems with Docker Compose8.2.3 Externalizing local agent microservices8.3 Considering advanced deployment strategies8.3.1 Choosing a runtime: Edge, API, or event-driven8.3.2 The three “wires” of communication8.3.3 Practical multi-agent topologies that adapt well8.3.4 State, memory, and idempotency8.3.5 Release engineering for agents (prompts, tools, models)8.3.6 Observability matters8.3.7 Reliability patterns: Timeouts, fallbacks, and budgets8.3.8 Cost control and model routing8.4 Security, safety, and governance in production8.4.1 A quick threat model for agentic systems8.4.2 Identity and access for people, services, and agents8.4.3 Secrets and configuration management8.4.4 Tool safety: Sandboxing and egress control8.4.5 Prompt-injection and data-exfiltration defenses8.4.6 Safety and policy enforcement8.5 Exercises
9 Understanding the agentic loop
9.1 Peeling back the three agentic loop layers9.1.1 Layer 1: The inner loop (sense-plan-act-learn)9.1.2 Layer 2: The task loop9.1.3 Layer 3: The meta loop9.2 Layer 2: Looping with a deep research agent9.2.1 Creating the initial state and plan9.2.2 Adding the tools9.2.3 Understanding iteration body output9.2.4 The termination gate9.2.5 Coding the deep research loop9.2.6 Synthesizing the final output9.2.7 When to use an agentic loop9.2.8 Building a repetitive task loop agent9.3 Layer 3: Multi-agent orchestration loops9.4 Building collaborative agentic loops9.5 Exercises
10 Exploring the cognitive agent that thinks, monitors, and adapts
10.1 Understanding agent cognition and metacognition as engineering concepts10.1.1 The five failure modes of capable-but-not-cognitive agents10.1.2 From reasoning primitives to cognitive architecture10.1.3 Defining cognition for agents10.1.4 Defining metacognition for agents10.1.5 Three theoretical foundations10.2 Mapping the mind into a cognitive agent architecture10.2.1 Architecture overview10.2.2 The cognitive workspace10.2.3 The perception module10.2.4 The planning module10.2.5 The execution module10.2.6 The evaluation module10.2.7 The attention module10.2.8 The memory module and the MCP memory server10.3 Building and running the cognitive agent10.3.1 The cognitive loop10.3.2 A complete cognitive agent with MCP10.3.3 Walkthrough: Watching the cognitive cycle in action10.3.4 Confidence-gated execution10.3.5 Stagnation detection and strategy pivoting10.3.6 Knowledge boundary awareness10.3.7 Emergent behaviors10.4 Measuring cognitive capability and looking ahead10.4.1 Cognitive efficiency metrics10.4.2 Before and after: Measuring the effect10.4.3 The road to more general agents10.5 Exercises
11 Tips for building agentic systems
11.1 Field-tested tips organized by the five agentic layers11.1.1 The core layer: Persona11.1.2 Tools and agent actions11.1.3 Reasoning and planning11.1.4 Knowledge and memory11.1.5 Evaluation and feedback11.2 Tips for building a customer support agent11.3 Tips for building a RAG agent system11.4 Tips for building a deep research agent system
appendix A Setting up the sample code repository
A.1 Cloning the repositoryA.2 Creating a Python environmentA.3 Installing dependencies and configuring the environmentA.3.1 Path A: Using the VS Code debuggerA.3.2 Path B: Installing manually with pipA.3.3 Configuring the OpenAI API keyA.4 Running the sample codeA.4.1 Running a sampleA.4.2 Troubleshooting common problemsA.4.3 Keeping your setup healthy
appendix B Node.js setup for local MCP servers
B.1 Installing Node.jsB.1.1 Installing Node.js on WindowsB.1.2 Installing Node.js on macOSB.1.3 Installing Node.js on Linux or WSLB.2 Verifying your Node and npx installationB.2.1 Checking the installed versionsB.2.2 How npx finds and caches packagesB.3 Running an MCP server with npxB.3.1 Anatomy of the npx commandB.3.2 Running the filesystem MCP serverB.3.3 Wiring the server into an MCP clientB.4 Troubleshooting and keeping Node healthyB.4.1 Common issuesB.4.2 Clearing the npx cacheB.4.3 Updating Node

Content preview from AI Agents in Action, Second Edition

7 Building robust agents with evaluation and feedback

This chapter covers

Introducing agent evaluation and feedback
Implementing test-driven agent development
Employing grounding, critic, and evaluation agents
Using Phoenix for evaluation and feedback

Evaluation and feedback provide the discipline that makes agent robustness measurable and improvable. They do not produce robustness on their own; a poorly architected agent will fail in ways that no evaluation suite can fix. What evaluation and feedback give you is visibility into how the system actually behaves and a mechanism for iterating toward better behavior over time.

Agent evaluation takes many forms, from benchmark and red team testing to grounding checks and agents that evaluate ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Publisher Resources

ISBN: 9781633434530Publisher Support Other Publisher Website Purchase Link

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

AI Agents in Action, Second Edition

by Micheal Lanham