AI AGENTS IN PRACTICE
Build. Deploy. Scale. Federate. Trust.
AI Agents in Practice is a four-book engineering series for developers, software engineers, and technical professionals who want to move beyond AI demonstrations and build production-ready agent systems that can operate reliably in the real world.
Rather than treating AI agents as isolated prompts or experimental prototypes, the series follows one continuous engineering journey: NexaCore, a fictional software company whose support agent evolves from a small Python prototype into a sophisticated, distributed, multi-agent federation operating across organisational boundaries.
Each book builds directly on the previous one. The result is a structured path through the architecture, engineering, operations, security, coordination, and governance challenges that emerge as AI systems become more capable and more consequential.
BOOK 1 — AI AGENTS: FOUNDATIONS
Start by understanding how AI agents actually work and build one from the ground up.
You will move from the fundamentals of agent architecture to working Python implementations covering reasoning loops, tools, short-term and long-term memory, RAG, prompt engineering, planning, error handling, observability, testing, security, prompt injection, and deployment.
The goal is not simply to understand what an AI agent is. You will build one that can be deployed.
BOOK 2 — AI AGENTS: BUILDING & DEPLOYING
Once an agent works, production introduces a new set of problems.
Learn how to scale agent execution with asynchronous processing and batch pipelines, control token and API costs, implement semantic caching, design multi-agent architectures, coordinate specialist agents, manage shared memory, generate and version tools, integrate human-in-the-loop workflows, trace production systems, run A/B tests, and design disaster recovery strategies.
This is the transition from a working agent to a production system capable of handling real operational pressure.
BOOK 3 — AI AGENTS: MULTI-AGENT SYSTEMS
What happens when agents must cooperate beyond the boundaries of a single organisation?
This volume introduces federated agent networks, where independent organisations operate their own agents but still need to coordinate securely.
Explore identity and trust, data isolation, federated tool registries, negotiation protocols, consensus, long-horizon planning, simulation, partial failure, governance, audit trails, rate limiting, fairness, security, emergent behaviour, and self-organising networks.
The focus shifts from building one organisation's AI system to engineering cooperation between independent systems.
BOOK 4 — AI AGENTS: ADVANCED ENGINEERING
The final volume addresses the hardest question: how do you make highly capable autonomous agents trustworthy?
Learn how to engineer long-running agents, evaluate systems at scale, design failure recovery and self-correction, establish capability boundaries, red-team agent systems, work with preference learning, detect reward hacking and specification gaming, implement constitutional constraints, fine-tune for specialised domains, build advanced RAG with provenance validation, develop multimodal agents, optimise latency and cost, and establish institutional governance and engineering ethics.
Across four books, the series follows a clear progression:
FOUNDATIONS → PRODUCTION → FEDERATION → TRUST
Designed around a continuous engineering project rather than disconnected examples, AI Agents in Practice shows not only how agent systems are built, but why particular architectural decisions become necessary as those systems grow.
Whether you are beginning your journey into agent engineering or already operating production AI systems, this series provides a practical framework for understanding the engineering problems that appear at every stage of the journey.
Build the agent. Deploy it. Scale it. Connect it. And learn what it takes to make it trustworthy.
