An AI agent is an LLM in a loop with tools. That's the simple definition. The complex part is everything that makes it work in production: reasoning patterns, tool design, memory systems, evaluation, safety, and the dozen subtle decisions that separate a demo from a system that holds up under real traffic.
This section is 45 pages on building agents that work. From ReAct loops to multi-agent orchestration, from tool design to production observability. The goal: give you the mental models and specific patterns you need to build something real.
What agents are, when to use them, the architecture map.
ReAct, planning, reflection, self-correction.
Tool design, descriptions, error handling, budgets.
Short-term, long-term, episodic, procedural.
Orchestration patterns, handoffs, debate.
Task completion, trajectory eval, regression testing.
Observability, cost, latency, safety, human-in-loop.
Research, coding, support, data, browser, email agents.
Claude Agent SDK, LangGraph, CrewAI, AutoGen.
Start at Foundations. If you already know what agents are, jump to Loops then Tools. Evaluation and Production are where naive agents fall apart at scale.