Academy / Generative AI Engineering

Path 3 · Course 2 · For developers and technical teams

AI Agents & Agentic Workflows

Design agents that plan, use tools and take actions: workflows vs agents, tool use and MCP, the five workflow patterns, the agent loop, guardrails and human approval, and evaluating and operating agents — with working Python.

📘 6 modules⏱ About 3.5 hours✅ Knowledge check per module🎓 Final assessment & certificate🎯 Basic Python

0 of 6 modules complete

Start Module 1 Final assessment

Modules

  1. 1

    What makes an AI agent

    Workflows vs agents, the augmented LLM, when (not) to use agents, and the autonomy spectrum.

    25 min
  2. 2

    Tool use and the Model Context Protocol

    How function calling works, defining tools, a full round of tool use in Python, designing good tools, and MCP.

    35 min
  3. 3

    Workflow patterns

    Prompt chaining, routing, parallelisation, orchestrator–workers and evaluator–optimiser — with code for routing and evaluator–optimiser.

    30 min
  4. 4

    Building an agent loop

    A bounded agent loop in Python, state and memory, managing context, planning, multi-agent systems, skills and frameworks.

    35 min
  5. 5

    Guardrails and human-in-the-loop

    Agent failure modes, guardrails by design, approval gates in code, escalation, failure handling and prompt injection.

    30 min
  6. 6

    Evaluating and operating agents

    Agent metrics, building evaluations with mock tools, trajectory grading, repeated runs, observability, staged roll-out and governance.

    30 min