Loading johnnynguyen.cloud portal
Microsoft Azure · initializing blades

Foundry

AI-103 Agent Lab · v0.4.0

An agent-building trainer for AI-103, the brand-new Azure AI Apps and Agents Developer exam. Drag models, knowledge, tools, and guardrails onto a Microsoft Foundry style canvas, wire them together, and check your build against exam scenarios.

The actual lab, running live: the multi-agent orchestration template loaded on the Agent Builder canvas. Click anywhere to open it full size.

Why it exists

A Microsoft Cloud Solution Architect asked me to build it. AI-103 is so new that real prep resources barely exist: it replaces AI-102 in June 2026 and pivots hard into building agents code-first in Microsoft Foundry. So I turned the official study guide into a place where you practice the exam's real skill, assembling working agent architectures, instead of memorizing service names.

It runs on the same React and TypeScript stack as this portal, with a hand-rolled SVG canvas inside an M365 Copilot style shell. The whole thing is a static site on Azure Static Web Apps, billed to my Azure for Students credit.

What you can do

  • Assemble agents from models, knowledge sources, tools, and guardrails on a live canvas
  • Wire components with typed links: data, grounding, identity, event, and monitoring
  • Pick an exam-style scenario challenge, then check whether your agent design solves it
  • Load reference templates: RAG chat, multi-agent orchestration, document pipelines, and workflows
  • Ask for an AI review and get your design read back in plain language
  • Drill with quizzes, case studies, performance-based questions, and flashcards too

Built from the source

Every scenario, quiz, and flashcard is grounded in Microsoft's official AI-103 study guide and weighted the way the published blueprint is, so the domains you see most are the domains that count most. The AI reviewer uses my own Azure OpenAI resource, the same one that powers Johnny Copilot on this portal.