Microsoft Foundry: End-to-End Workshop¶
A hands-on coding workshop for building enterprise AI agents and apps on Microsoft Foundry (Azure AI Foundry).
Welcome! You'll go from your first model call to a grounded, tool-using, evaluated, observable agent — and finish by combining it all into a capstone. Every lab is a self-contained Jupyter notebook that also renders as a page on this site, so you can read along or run it yourself against your own Foundry project.
One project, az login, and you're off
No hub-and-spoke Bicep, no APIM. The whole workshop runs against one Foundry
project authenticated with DefaultAzureCredential. Start with
Setup.
Read the platform primer first¶
Three short, code-free pages give you the mental model before lab 1:
| Page | What it covers |
|---|---|
| What is Foundry | The unified PaaS, the two portals, what's built in. |
| Architecture & projects | Accounts, projects, connections, model deployments. |
| Control plane & governance | Provisioning, regions, RBAC, gateways, Azure Policy. |
The labs¶
Work through them in order — each builds on the last, and the client bootstrap from M1 is reused everywhere.
| Module | You'll learn to… |
|---|---|
| M1 · First inference | Call models with AIProjectClient — chat, embeddings, streaming, the Responses API. |
| M2 · Your first agent | Create a versioned prompt agent and invoke it via agent_reference. |
| M3 · Tools & function calling | Attach the Code Interpreter and your own custom function tools. |
| M4 · Grounding / RAG (Foundry IQ) | Ingest a corpus into Azure AI Search and ground an agent with citations. |
| M5 · MCP tools | Connect an agent to a Model Context Protocol tool catalog. |
| M6 · Agent memory | Give an agent cross-turn memory. |
| M7 · Multi-agent orchestration | Route between specialist agents with the Agent Framework. |
| M8 · Deep research | Run an agentic research loop with cited synthesis. |
| M9 · Evaluation | Score quality with built-in, agent, and custom evaluators. |
| M10 · Observability & tracing | Trace agents with OpenTelemetry → App Insights; evaluate continuously. |
| M11 · Guardrails | Stack Prompt Shields, PII detection, and a custom blocklist. |
| M12 · Red teaming | Run automated adversarial scans against your agent. |
| M13 · Human-in-the-loop & REST | Pause for approval; invoke an agent over raw REST. |
| M14 · Fine-tuning | Distill a teacher model into a small student (LoRA/PEFT). |
| M15 · Capstone | Combine everything into one agent — and where to go next. |
Suggested agenda¶
This is comfortably a two-day workshop; compress to one long day by treating M8 / M12 / M14 as "read & skim".
| Block | Modules |
|---|---|
| Day 1 AM — Foundations | Platform primer · M1 · M2 · M3 |
| Day 1 PM — Knowledge & tools | M4 · M5 · M6 · M7 |
| Day 2 AM — Quality & safety | M8 · M9 · M10 · M11 |
| Day 2 PM — Hardening & capstone | M12 · M13 · M14 · M15 |
How to use this site¶
- Do Setup first — ~15 minutes, before the workshop.
- Read the Concepts for the through-line that ties the labs together.
- Work the labs in order. Each ends with a 🧪 Your turn exercise.
Run it yourself, anytime
Code cells render with an "Expected output" note in the surrounding text, so the site is useful even without Azure. To actually run a lab, use the Download link at the top of its page and point it at your own Foundry project.
Ready? → Set up your environment