Microsoft Agent Framework Training Day

[TallesValiatti]

Introduction

Hello everyone,

Together with Carlos Machel, and representing AzureBrasil.Cloud, we delivered a full-day, hands-on training session focused on the Microsoft Agent Framework, guiding participants in the practical development of intelligent agents with .NET and Azure.

Carlos and I served as the main instructors, and we were supported by the exceptional AzureBrasil.Cloud team: Rafael dos Santos, Luiz de Assis, Otavio Namaike, Talles Werner, Felipe Klafke, Felipe Farias, Charlison Oliveira, as well as our partner Renyo Borges, who all contributed to making the training day smooth, productive, and impactful.

The training took place at Coco Bambu, in Vitória/ES, Brazil, on November 28th, 2025, running from 9:00 AM to 6:00 PM.

Throughout the day, we covered and practiced several advanced features of the agent ecosystem, including:

• Function Calling
• Context Provider
• Integration with Azure AI Search – Knowledge Agent
• Multi-Agents
• Workflows
• Integrations with MCPs
• Autonomous Agent
• Fine-tuning with Microsoft Foundry

All training materials, including the code samples used during the session, are available in the repository:

https://github.com/AzureBrasil-cloud/ai-training-session-microsoft-agent-framework

This training provided an in-depth look at how the Microsoft Agent Framework can be applied in real enterprise scenarios, exploring modern architectures and advanced integrations across the Azure ecosystem.

Features and Screenshots

Technical Details

• C# (.NET 10)
• Microsoft Agent Framework
• Microsoft Foundry
• Azure AI Search (including Knowledge Agent integration)
• MCPs (Model Context Protocol) integrations
• Microsoft Foundry (for fine-tuning workflows)
• Docker (local development environment)
• VS Code / Visual Studio / Rider (development tools)
• GitHub (source control and collaboration)

Challenges and Solutions

Problems Encountered

• Integrating multiple agents and ensuring consistent coordination between them.
• Handling context propagation across workflows and tool calls.
• Setting up the Knowledge Agent with Azure AI Search and aligning index structure with agent expectations.
• Managing MCP integrations and validating tool responses.
• Ensuring smooth execution of function calling with complex payloads.

Solutions Implemented

• Structured each agent with clear responsibilities to reduce overlap and simplify coordination.
• Introduced custom context providers to maintain consistent state across workflows.
• Adjusted Azure AI Search indexing strategy and added normalization steps for better retrieval quality.
• Implemented validation layers and response handlers for MCP tool interactions.
• Standardized function schemas and added lightweight input/output validation to prevent execution errors.

[carlosmachel]

It was a pleasure delivering this training with you and with the support of AzureBrasil.cloud. I'm excited for the next training sessions.

[rpbs]

I was invited by @TallesValiatti to help on the support for this event, the content about MAF was unique and the use case for the demos presented reflect on real situations, and once again AzureBrasil is one step ahead when it comes about Azure and AI.

Specially thanks for @TallesValiatti and @carlosmachel