By concept

A learning path through agentic systems — each concept links to the demos that show it in action.

Foundations

• Agent fundamentals

  3

  The smallest useful agent: a model, an instruction, a loop.

  See demos →

• Tool use & function calling

  6

  Giving a model hands — letting it call functions and APIs.

  See demos →

Reasoning & orchestration

• Workflows & orchestration

  3

  Routing, parallel steps, and deterministic graphs around an LLM.

  See demos →

• Multi-agent systems

  4

  Several specialised agents collaborating on one goal.

  See demos →

Tools & protocols

• MCP (Model Context Protocol)

  2

  A standard way to plug tools and data into any agent.

  See demos →

• A2A & agent registry

  2

  Agents discovering and calling each other, no hardcoded URLs.

  See demos →

Memory & knowledge

• Agent memory

  3

  Short- and long-term memory so agents remember across turns.

  See demos →

• RAG & data agents

  2

  Grounding answers in retrieved documents and live data.

  See demos →

Interfaces

• Browser & computer use

  4

  Agents that drive a real browser or desktop to get things done.

  See demos →

• Voice agents

  2

  Real-time, spoken conversations with an agent.

  See demos →

• Accessibility agents

  2

  Agents built to make the web and apps usable for everyone.

  See demos →

• Multimodal & creative

  3

  Agents that see, hear, generate images, music, and video.

  See demos →

Quality

• Evaluation & quality

  2

  Measuring whether an agent actually works — and stays working.

  See demos →

• Agent skills

  1

  Packaging reusable capabilities an agent can load on demand.

  See demos →