No. 278 of 500 No. 31 in Generative AI & LLMs

Building Applications with AI Agents

Designing and Implementing Multiagent Systems

By Michael Albada

Print length
352 pages
Publisher
O'Reilly Media
Reading level
Intermediate
ISBN-13
978-1098176501
Published
October 21, 2025
Written for
Software engineers and technical product leads building their first LLM agent or multiagent application.

Synopsis

Michael Albada, a machine learning engineer who has built agent systems for security work at Microsoft, explains how foundation models become agents: language models that emit function calls, use tools, retrieve knowledge, keep memory and chain several model calls to finish tasks no single prompt can handle. Opening chapters distinguish fixed workflows from agents that choose their own next steps, introduce the core components of an agent system and set out the design trade-offs among them, treating agency as a matter of degree rather than an on-off property.

Later chapters cover tool use, which Albada describes as the building blocks that let agents retrieve information and act on their environment; orchestration, including context engineering, tool selection and planning topologies; knowledge and memory, separating retrieval-augmented generation from an agent's record of its own history; coordination among multiple agents; and protecting agentic systems against threats such as prompt injection and memory poisoning. Rather than tie the book to one library, he implements the same scenarios in LangGraph, LangChain and AutoGen so readers can compare trade-offs directly. Engineers gain patterns for deciding when an agent is warranted, how much autonomy to grant and how to move a prototype toward production code.

Scorecard

Editorial score

How it scored

Editorial score: 66.8 / 1005 weighted criteria

  • Influence30%

    16.5/30
  • Authority25%

    16.3/25
  • Practical value20%

    15.0/20
  • Currency15%

    12.0/15
  • Readability10%

    7.0/10
  • Editorial score

    66.8/100

How the rank was computed

Rank No. 278 of 500 comes from a composite of 53.6: observable signals 23.5 of 55 points plus editorial review 30.1 of 45 points.

All 11 inputs
InputPoints
Observable signals23.5/55
Reach (Amazon ratings volume)6.6/16.5
Consensus across best-of lists8.3/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted4.4/8.3
Amazon sales rank1.4/5.5
Recency2.8/2.8
Editorial review30.1/45
Author authority7.3/11.3
Influence7.4/13.5
Practical value6.8/9.0
Readability3.2/4.5
Currency5.4/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A sound, principles-first introduction to designing LLM agents and multiagent systems for engineers new to agentic software.

Strengths

  • Principles that outlast specific framework changes
  • Same scenarios compared across three frameworks
  • Author's production experience building security agents

Keep in mind

Framework-specific code may date quickly as LangGraph, LangChain and AutoGen evolve, and readers already experienced with agents will find much of it introductory.

Key takeaways

  1. Agency is a spectrum: more guardrails mean less autonomy, and constrained, workflow-like systems can deliver reliable value before fully autonomous agents.
  2. Tools are the building blocks that let agents retrieve information and act on their environment, so careful tool design is central to agent reliability.
  3. Knowledge retrieval and memory solve different problems: retrieval brings in outside facts, while memory preserves an agent's own history of exchanges and tool outputs.
  4. Agents introduce new attack surfaces, including prompt injection and memory poisoning in multiagent workflows, so security must be designed in from the start.

Notable facts

  • Victor Dibia included it in his list of top books on AI agents, calling it "a solid introductory resource for folks new to AI agents." Source (opens in a new tab)

About the author

Michael Albada

Michael Albada is a machine learning engineer with nine years of experience designing, building and deploying large-scale machine learning systems at Uber, ServiceNow and Microsoft. His work spans recommendation systems, geospatial modeling, cybersecurity, natural language processing and large language models, including multiagent systems for security tasks such as threat hunting at Microsoft. This is his first book.

Questions

What does Building Applications with AI Agents cover?
It explains how to design and implement software agents built on foundation models, covering how agents use tools, orchestrate steps, retrieve knowledge, maintain memory and coordinate in multiagent systems, plus how to protect them from attacks. Michael Albada emphasizes design principles and trade-offs over any single library, implementing the same scenarios in LangGraph, LangChain and AutoGen.
Is Building Applications with AI Agents good for beginners?
It suits developers who are new to agents but comfortable with Python and the basics of large language models; Victor Dibia calls it a solid introductory resource. Readers with no programming background will struggle with the code, and engineers who already run agents in production may find most concepts familiar and should look for more advanced material.
Is Building Applications with AI Agents still current in 2026?
Its principles remain sound: when to use an agent, how much autonomy to grant, how to design tools and how to structure multiagent work. Its framework-specific code ages faster, because agent libraries change quickly. Readers should treat the code as illustrations of patterns and check current documentation for LangGraph, LangChain or AutoGen before adapting it.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.