Module 1 · Introduction to AI Agents

OBJECTIVE: understand what an AI agent is, how it works step by step, when you need one (and when a simple workflow is better), and which tools are used to build agents. WHAT WE COVER: • What an agent is: an LLM that can decide, use tools and act in a loop until a goal is done • Chatbot vs workflow vs agent • The agent loop and the ReAct pattern (think → act → observe) • Types of agents: reflex, goal-based, utility-based, learning and tool-using agents • The 5 parts of an agent: model, tools, memory, planning, guardrails • Workflow patterns: chaining, routing, parallel, orchestrator–workers, evaluator–optimizer • Autonomy levels and human-in-the-loop • The framework landscape: LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, LlamaIndex — and how to choose AFTER THIS MODULE YOU CAN: • Explain how an agent works and draw its loop • Decide if a problem needs an agent, a workflow or just one prompt • Build a tiny agent from scratch in plain Python, without any framework BUILDS ON: tool calling (LLM module, Project 1) and RAG (LLM module, Project 2) — not repeated here. MINI EXERCISE: a 60-line ReAct agent with 2 tools (calculator and web search) written from scratch.