Agents

What you'll learn

1. 1

   Workflows vs Agents

   The five patterns you'll use most, and when NOT to build an agent.

   • Two kinds of agentic systems
   • Name that pattern
   • When to use a full agent
   • Build a support-ticket router

   Take-home template: Workflow vs Agent Decision Checklist

2. 2

   The Agent Loop

   Build the core loop behind every agent, with real guardrails.

   • An agent is a loop
   • Watch: an agent think, act and observe
   • One iteration
   • Write an agent loop
   • Agent stuck in a loop

   Take-home template: Production Agent Loop

3. 3

   Memory for Assistants & Agents

   Short-term context, long-term memory, and how to pick what to remember.

   • Models don't remember. Systems do.
   • Rank memories
   • Where should it live?
   • Right to be forgotten
   • Samsung's source-code leak

   Take-home template: Long-Term Memory Store (SQLite)

4. 4

   Multi-Agent Systems

   Orchestrators, subagents and handoffs, and why more agents isn't automatically better.

   • When one agent isn't enough
   • Fan out work in parallel
   • Orchestrator workflow
   • Is it worth it?

   Take-home template: Orchestrator-Workers Template

5. 5

   MCP: Model Context Protocol

   The open standard for plugging tools and data into AI apps.

   • USB-C for AI tools
   • MCP vocabulary
   • Write the core of an MCP server
   • Security check

   Take-home template: Your First MCP Server (Python)