Introduction
Welcome to the Microsoft Agent Framework Workshop — a hands-on guide to building production-ready AI agents in Python.
What Is Microsoft Agent Framework?
Microsoft Agent Framework is an open-source Python SDK for building, orchestrating, and hosting AI agents. It provides first-class support for:
- Tool calling — attach any Python function as an agent tool
- MCP (Model Context Protocol) — connect agents to live external data sources
- Multi-turn conversations — built-in history and memory providers
- Multi-agent workflows — orchestrate agents into pipelines
- Hosting — deploy agents via Azure Functions, A2A, OpenAI-compatible, or AG-UI endpoints
What You Will Build
This workshop guides you through building OpsAgent — an AI-powered operations and engineering assistant — from a blank Python project to a fully hosted HTTP endpoint.
Each module builds directly on the previous one. By the end you will have a single agent that uses tools, retrieves live documentation via MCP, remembers conversation history, participates in multi-agent workflows, runs inside multiple chat UIs, and is deployed as a serverless Azure Function.
Who This Workshop Is For
| Background | What You Will Learn |
|---|---|
| Python developers new to AI agents | How to build and run your first agent |
| Cloud / DevOps engineers | How to wire tools, MCP, and host agents |
| AI/ML engineers | How to compose multi-agent workflows |
No prior experience with Azure or AI agents is required — just Python 3.10+ and a GitHub account.
Workshop Structure
The workshop is split into 10 modules under Getting Started:
| Module | Topic |
|---|---|
| 1 | Environment Setup |
| 2 | GitHub Models Connection |
| 3 | Microsoft Agent Framework Agents |
| 4 | Tool Calling |
| 5 | MCP Integration |
| 6 | Multi-Turn Conversations |
| 7 | Memory & Persistence |
| 8 | Workflows |
| 9 | Chat User Interface |
| 10 | Host Agent |
Each module has a corresponding test file in lab/ so you can validate each step independently before moving on.
Key Technologies
| Technology | Role |
|---|---|
| Microsoft Agent Framework | Core agent SDK |
| GitHub Models | Free LLM inference (GPT-4o mini) |
| MCP — Model Context Protocol | Live data retrieval |
| Azure Functions | Serverless agent hosting |
| Azurite | Local Azure Storage emulator |
Before You Start
Head to Module 1 — Environment Setup to configure your local development environment.
!!! tip “GitHub Token” You will need a GitHub Personal Access Token with Models permission enabled. Create one at github.com/settings/tokens.