Pro-Code Agent Demo Site

Customer-facing walkthrough showing how to build a pro-code agent that:

• Plugs into the customer's own agentic harness (no Foundry Agent Service / Copilot Studio)
• Calls models hosted in Azure AI Foundry via the AOAI-compatible endpoint
• Uses Microsoft Entra Agent ID for first-class agent identity
• Uses Microsoft Purview to classify and gate sensitive content on the way in and out
• Provides equivalent C# and Python samples for every concept

Running the site

It's a single static page — no build step. Pick whichever fits the demo machine:

# Recommended: bundled PowerShell server (no Python / Node required).
# Opens the browser automatically.
.\serve.ps1                  # http://localhost:8080
.\serve.ps1 -Port 9000       # custom port
.\serve.ps1 -NoBrowser       # don't auto-open the browser

# Or, if you have them:
python -m http.server 8080
npx serve .

# Or just open the file (works in Edge/Chromium, not Firefox)
ii index.html

The samples are pulled into the page with fetch(), so a local server is the safest demo path — file:// is blocked by Firefox and some browser policies.

Layout

.
├── index.html                 # Overview / docs page
├── wizard.html                # Interactive 46-step build-up wizard
├── serve.ps1                  # One-command static server (PowerShell)
├── css/
│   ├── styles.css             # Shared theme
│   └── wizard.css             # Wizard-specific layout
├── js/
│   ├── app.js                 # Docs page interactions
│   ├── wizard.js              # Wizard runtime (parser, render, nav)
│   └── wizard-data.js         # Per-step titles + explanations
└── samples/
    ├── csharp/                # Stand-alone runnable samples
    │   ├── BasicAgent.cs
    │   ├── EntraAgentIdentity.cs
    │   ├── PurviewClassification.cs
    │   └── EndToEndAgent.cs
    ├── python/                # Stand-alone runnable samples
    │   ├── basic_agent.py
    │   ├── entra_agent_identity.py
    │   ├── purview_classification.py
    │   └── end_to_end_agent.py
    └── wizard/                # Tagged sources consumed by wizard.js
        ├── 01_basic.{cs,py}
        ├── 02_entra.{cs,py}
        ├── 03_purview.{cs,py}
        └── 04_e2e.{cs,py}

Files under samples/wizard/ are templates — they contain trailing // @N / # @N step tags that the wizard parses. Don't run them directly; the runnable equivalents live under samples/csharp/ and samples/python/.

Adapting for a customer

Edit four constants and you're talking to a real environment:

Where	Replace	With
Sample files (both languages)	https://contoso-foundry.services.ai.azure.com/…	Customer's Foundry project endpoint
Sample files (both languages)	gpt-4o-mini	The customer's deployment name
entra_agent_identity.*	contoso.onmicrosoft.com	Customer's tenant domain
entra_agent_identity.*	00000000-0000-0000-0000-000000000000	Object ID of the human / team that owns the agent

Tenant used in the live demo

The walkthrough is wired against models deployed in the admin@contoso.com tenant. Sign in with that account when you need to exercise the live calls.