Skip to content

Defaults & paths ​

The default stack ​

The workshop examples all use one default stack so everyone can follow along. You are welcome to adapt them, see Choose your path below.

LayerDefaultWhy
Source control & workflowGitHubRepo, pull requests, and CI/CD (Continuous Integration / Continuous Deployment) in one place.
AI ecosystemAzure + Microsoft Foundry (formerly Azure AI Foundry)The default place we run and manage the model.
LanguagePythonThe most common language for agent tooling and the lab samples.
Terminal shellBashAll lab commands are written for bash. Git Bash, WSL, macOS Terminal, or any Linux shell will work.
Workshop siteVitePressThis site you are reading.

"Default" means example, not requirement

Every concept (context engineering, memory, tools, MCP, HITL, A2A) is portable. The examples happen to use the stack above; the ideas work anywhere.

Choose your path ​

Pick the path that fits your goals and setup today. You can switch paths at any time.

PathBest whenWhat you do
Default pathYou want hands-on practice and have the stack readyRun every lab exactly as written.
Adapt to your own stackYou already use a different language/cloud/frameworkMap each lab concept onto your tools; use our steps as a reference.
Observe conceptuallyYour accounts aren't ready, or you prefer to watchRead along, run nothing, and use the diagrams and checkpoints to follow the ideas.

No wrong choice

If a lab isn't working on your machine, switch to observe for that hour and rejoin the default path at the next checkpoint. You will not fall behind.

Guidance by experience level ​

This day is designed for software developers who may be new to agentic AI. Use the row that matches you, and feel free to move between them.

LevelYou probably…Focus on…
BeginnerWrite code, but haven't built with AI agentsThe plain-language definitions and the default path. Don't skip the "Why it matters" sections.
IntermediateHave tried a chatbot or one tool callComparing the safe patterns (typed tools, memory boundaries, HITL) to what you've done before.
AdvancedHave built agents alreadyThe stretch tasks in each segment and the failure modes, where these patterns break and why.

What we will not cover ​

  • A framework-by-framework comparison; we go deep on the default stack and keep concepts portable.
  • Choosing a model provider or doing cost/latency benchmarking.