Sixteen topics from the PDF outline (AI Agents, pages 177–264 — the chapter before MCP) in the same reference deck as LLMs, Prompt engineering, Fine-tuning, RAG, and Context engineering—from definitions and building blocks through ReAct, protocols, optimization, and deployment.
Chapter divider plus the opening narrative: manual LLM iteration vs autonomous agents that search, filter, summarize, and format end-to-end (PDF 177–180).
Brain vs fresh knowledge vs the decision-maker: how an agent plans and acts using an LLM plus tools and retrieval (PDF 181–182, shared with the next topic).
Role, focus, tools (including custom tools and MCP), cooperation, guardrails, and memory—through the deck’s CrewAI-oriented examples (PDF 182–196; 182 shared with topic 2).
Semantic, episodic, and procedural angles on long-term memory—and how typed memory supports tutoring-style personalization (PDF 196–197; shared boundaries with building blocks and Importance).
Stateless loops vs recall across iterations—and why memory is a systems problem, not a model property (PDF 197–200; tail of 200 shared with design patterns).
Reflection, tool use, ReAct, planning, and multi-agent collaboration—framework-ready patterns with CrewAI callouts where relevant (PDF 200–206; shared boundaries on 200 and 206).
Notebook-style walkthrough: minimal agent class, a structured ReAct protocol prompt, manual tool injection, and automated loops with LiteLLM (PDF 206–231; 206 header shared with design patterns).
Maturity framing for how much autonomy, tooling, and orchestration a system exhibits (PDF 232–235).
Glossary-style tour of vocabulary you will see in papers, vendors, and codebases (PDF 236–240).
Stack view of how models, tools, orchestration, and product interfaces layer into shipped agents (PDF 241–242).
Coordination motifs when several agents divide work, debate, or review each other (PDF 243–245).
How agents discover and message peers in the emerging A2A framing from the deck (PDF 246–248).
User-facing interaction contracts—streaming, approvals, and UI affordances for agentic apps (PDF 249–252).
How MCP, A2A, AG-UI, and adjacent standards relate in one picture from the course deck (PDF 253–255).
Tracing, evaluation, and iterative improvement loops for agent runs using Opik from the deck’s examples (PDF 256–260).
From prototype notebooks to production placement, scaling, and governance considerations (PDF 261–264; MCP chapter follows on 265).