Librería Samer Atenea
Kálamo Books
Librería Elías (Asturias)
Librería Kolima (Madrid)
Librería Proteo (Málaga)
Go beyond prompts and prototypes with 15 proven patterns for engineering reliable, secure, and production-ready AI agentsKey Features:- Master durable agent harness patterns for tools, context, memory, permissions, orchestration, multi-agent systems, and more- Compare real production approaches in TypeScript and Python and examine how different agents solve the same architectural problems differently- Turn agent prototypes into production systems and apply each pattern to a practical defensive cybersecurity agentBook Description:Building an AI agent is easy. Building one that can operate reliably in production is a much harder engineering problem.Harness Patterns for Agentic AI with Claude Code and Hermes Agent goes beyond prompts and agent frameworks to reveal the architectural patterns that make production agentic systems work. By examining Claude Code and Hermes Agent side by side, you’ll discover 15 reusable harness patterns for controlling how agents use tools, manage context, coordinate tasks, enforce permissions, retain memory, recover from failures, and interact with external systems.Each pattern is explored through real TypeScript and Python implementations, followed by a practical comparison that explains the trade-offs and when to use each approach. You’ll then apply the pattern to a defensive cybersecurity agent, seeing how these architectural decisions behave under real operational pressure.From agent loops and tool contracts to permissions, context management, multi-agent coordination, observability, MCP, model routing, and structured outputs, you’ll develop a framework-independent mental model for engineering AI agents that are reliable, auditable, and ready for production.What You Will Learn:- Design robust agent loops and tool contracts that separate model reasoning from real-world actions- Build permission systems and safety guardrails that control destructive or sensitive agent operations- Manage growing context, persistent memory, state, and long-running agent interactions effectively- Orchestrate tools and multiple agents while balancing concurrency, coordination, and reliability- Instrument agentic systems with observability, debugging, hooks, events, and audit trailsWho this book is for:This book is for AI engineers, software engineers, architects, and technical leads who already understand the basics of LLMs and agents but need to engineer systems that can survive production workloads. It will also benefit platform engineers and security or compliance professionals responsible for evaluating agent architecture, permissions, auditability, and operational risk.Readers should be comfortable reading production code. Examples use both TypeScript and Python, but deep expertise in both languages is not required.Table of Contents- The Harness Paradigm- Tool Architecture and the Tool Contract- The Query / Agent Loop- Permission Systems and Safety Guardrails- Tool Orchestration and Execution- Context Management at Scale- Multi-Agent Coordination- Memory Systems and State Persistence- Observability and Debugging- Production Deployment Patterns- Hook / Event-Driven Automation- The Skill System Pattern- MCP Integration - Connecting Agents to the World- Model Routing and Provider Abstraction- Structured Output and Schema-Constrained Generation