Hosted Agents Workshop: Building & Deploying AI Agents on Microsoft Foundry

🎯 Workshop Overview

Welcome to this hands-on workshop where you'll learn to build, containerize, and deploy AI agents as hosted agents on Microsoft Foundry Agent Service. You'll use Python and the Microsoft Agent Framework to create production-ready agents that run as managed containerized services with autoscaling, observability, and enterprise-grade security.

What You'll Build

By the end of this workshop, you will have:

1. Lab 1: A fully functional hosted agent built with Microsoft Agent Framework, tested locally, containerized with Docker, and deployed to Microsoft Foundry Agent Service
2. Lab 2: Choose from multiple advanced agent patterns — a Web Search Agent, a RAG Agent, or a Multi-Agent Workflow — and publish your agent to channels

Learning Objectives

• Understand what hosted agents are and when to use them
• Set up a Microsoft Foundry project with the required resources
• Build an AI agent using the Microsoft Agent Framework (Python)
• Use the hosting adapter to test agents locally
• Containerize your agent with Docker and push to Azure Container Registry
• Deploy your agent to Microsoft Foundry Agent Service
• Manage the agent lifecycle — start, stop, scale, and update
• Monitor agents with OpenTelemetry tracing
• Publish agents to Teams, Web, and API channels

---

📋 Prerequisites

Required Knowledge

• Basic understanding of Python programming
• Familiarity with REST APIs
• Basic understanding of Docker and containers
• Understanding of basic AI/ML concepts

Required Tools

Tool	Download Link
Python 3.10+	https://www.python.org/downloads/
VS Code	https://code.visualstudio.com/
Git	https://git-scm.com/downloads
Docker Desktop	https://www.docker.com/products/docker-desktop/
Azure CLI	https://learn.microsoft.com/en-us/cli/azure/install-azure-cli

Required Access

• Azure subscription (create a free account)
• Microsoft Foundry access (create a project)

---

🗂️ Workshop Structure

This workshop is divided into two progressive labs, each with multiple sub-labs.

💡 Tip: Lab 1 is required before Lab 2. In Lab 2, you choose one (or more) agent patterns to explore.

Lab 1: Building & Deploying Your First Hosted Agent

Duration: 75–100 minutes

Sub-Lab	Description	Duration
1.1 Set Up Azure Resources	Create Foundry project, deploy models, set up ACR	15–20 min
1.2 Build Your Agent	Build a time-zone agent with Microsoft Agent Framework	15–20 min
1.3 Test Locally	Use the hosting adapter to test via REST	10–15 min
1.4 Containerize & Deploy	Docker build, push to ACR, deploy to Foundry	20–25 min
1.5 Manage & Monitor	Start, stop, scale, and observe your agent	15–20 min

Lab 2: Advanced Agent Patterns & Publishing

Duration: 60–90 minutes (choose your path)

Sub-Lab	Description	Duration
2.1 Web Search Agent	Build an agent that searches the web using Foundry tools	20–25 min
2.2 RAG Agent	Build a RAG agent with Azure AI Search integration	20–25 min
2.3 Multi-Agent Workflow	Orchestrate multiple agents in a workflow	25–30 min
2.4 Publish & Evaluate	Publish to channels and evaluate performance	15–20 min
2.5 Cleanup Resources	Delete all Azure resources	5–10 min

💡 In Lab 2, pick the agent pattern(s) that interest you most — you don't need to complete all three (2.1, 2.2, 2.3).

---

🏗️ Architecture Overview

What Are Hosted Agents?

Hosted agents are containerized agentic AI applications that run on Foundry Agent Service. Unlike prompt-based agents, you build hosted agents through code and deploy them as container images on Microsoft-managed pay-as-you-go infrastructure.

┌────────────────────────────────────────────────────────────────┐
│                    Microsoft Foundry                           │
│                                                                │
│  ┌──────────────┐    ┌──────────────────┐   ┌───────────────┐  │
│  │ Your Agent   │    │  Hosting Adapter │   │ Agent Service │  │
│  │ Code (MAF)   │───>│  (HTTP Server)   │──>│ (Managed)     │  │
│  └──────────────┘    └──────────────────┘   └───────────────┘  │
│         │                                          │           │
│         ▼                                          ▼           │
│  ┌──────────────┐                          ┌───────────────┐   │
│  │  Local Tools │                          │ Autoscaling   │   │
│  │  (Python)    │                          │ Observability │   │
│  └──────────────┘                          │ Conversations │   │
│                                            └───────────────┘   │
└────────────────────────────────────────────────────────────────┘

Key Components

Component	Description
Microsoft Agent Framework	Python/.NET SDK for building enterprise AI agents with type safety, checkpointing, and multi-agent orchestration
Hosting Adapter	Framework abstraction layer that exposes your agent as an HTTP service for local testing and deployment
Agent Service	Managed platform that handles provisioning, autoscaling, state management, identity, and observability

Supported Frameworks

Framework	Python	.NET
Microsoft Agent Framework	✅	✅
LangGraph	✅	❌
Custom code	✅	✅

This workshop uses Python with Microsoft Agent Framework.

---

🚀 Getting Started

1. Set up your environment — Install all prerequisites
2. Start Lab 1 — Build and deploy your first hosted agent
3. Continue to Lab 2 — Explore advanced patterns

---

📎 Additional Resources

• Hosted Agents Documentation
• Microsoft Agent Framework
• Foundry Samples — Hosted Agents
• Azure Developer CLI — AI Agent Extension
• Agent Development Lifecycle
• Foundry Pricing

📄 License

This workshop content is provided for educational purposes.

---

Ready to begin? Head over to the Setup Guide to get started!