Skip to content

Lab BAF0 - Prerequisites

In this lab you will set up your development environment to build, test, and deploy the custom engine agent you will develop using the Microsoft 365 Agents SDK and Agent Framework throughout this path.

In this lab you will learn how to:

  • Setup your Microsoft 365 environment
  • Install and configure Visual Studio Code with Microsoft 365 Agents Toolkit
  • Prepare your Azure environment to create required resources
  • Install required development tools

Disclaimer

These samples and labs are intended for instructive and demonstration purposes and are not intended for use in production. Do not put them into production without upgrading them to production quality.

Note

To install and run your own custom engine agent, you'll need a Microsoft 365 tenant where you have administrator permission. You won't need Microsoft 365 Copilot License to test your custom engine agent.

Exercise 1 : Setup Microsoft Teams

Step 1: Enable Teams custom application uploads

By default, end users can't upload applications directly; instead a Teams administrator needs to upload them into the enterprise app catalog. In this step you will ensure your tenant is set up for direct uploads by M365 Agents Toolkit.

1️⃣ Navigate to https://admin.microsoft.com/, which is the Microsoft 365 Admin Center.

2️⃣ In the left panel of the admin center, select Show all to open up the entire navigation. When the panel opens, select Teams to open the Microsoft Teams admin center.

3️⃣ In the left of the Microsoft Teams admin center, open the Teams apps accordion. Select Setup Policies, you will see a list of App setup policies. Then, select the Global (Org-wide default) policy.

4️⃣ Ensure the first switch, Upload custom apps is turned On.

5️⃣ Be sure to scroll down and select the Save button to persist your change.

The change can take up to 24 hours to take effect, but usually it's much faster.

Exercise 2: Setup Development Environment

You can complete these labs on a Windows, macOS, or Linux machine and you do need the ability to install the prerequisites. If you are not permitted to install applications on your computer, you'll need to find another machine (or virtual machine) to use.

Step 1: Install Visual Studio Code

1️⃣ Download and install Visual Studio Code from https://code.visualstudio.com/.

2️⃣ Launch Visual Studio Code after installation.

Step 2: Install .NET 9 SDK

The Microsoft 365 Agents SDK and Agent Framework require .NET 9 SDK to build and run the agent.

1️⃣ Download and install .NET 9 SDK from https://dotnet.microsoft.com/download/dotnet/9.0.

2️⃣ Verify the installation by opening a terminal and running:

dotnet --version

You should see version 9.0.x or higher.

Step 3: Install C# Dev Kit Extension

1️⃣ In Visual Studio Code, open the Extensions view by clicking on the Extensions icon in the Activity Bar on the side of the window or by pressing Ctrl+Shift+X (Windows/Linux) or Cmd+Shift+X (Mac).

2️⃣ Search for C# Dev Kit and click Install.

Step 4: Install Microsoft 365 Agents Toolkit Extension

1️⃣ In Visual Studio Code Extensions view, search for Microsoft 365 Agents Toolkit and click Install.

2️⃣ After installation, you should see the Microsoft 365 Agents Toolkit icon in the Activity Bar.

Step 5: Install Azure CLI

The Azure CLI is required to provision and manage Azure resources.

1️⃣ Download and install Azure CLI from https://learn.microsoft.com/cli/azure/install-azure-cli.

2️⃣ Verify the installation by opening a terminal and running:

az --version

3️⃣ Sign in to Azure:

az login

Step 6: Install DevTunnel

DevTunnel is required for local development and debugging. It creates a secure tunnel from the internet to your local machine.

Windows:

winget install Microsoft.DevTunnel

macOS/Linux:

curl -sL https://aka.ms/DevTunnelCliInstall | bash

Verify installation:

devtunnel --version

DevTunnel Alternative

DevTunnel is also included with Visual Studio 2022. If you have Visual Studio 2022 installed, you already have DevTunnel.

Exercise 3: Setup Azure Environment

To complete the exercises in this path, you'll need an Azure subscription to create Microsoft Foundry resources and deploy AI models.

Step 1: Get an Azure subscription

If you don't have an Azure subscription yet, you can activate an Azure free account that offers $200 in credits which can be used within the first 30 days on most Azure services.

Follow the steps to activate an Azure free account:

1️⃣ Navigate to Azure free account page and select Activate.

2️⃣ Login with an account of your choice, it's recommended to use the Microsoft 365 tenant account you would like to use in the exercises.

3️⃣ Check the boxes for Privacy Statement, then select Next.

4️⃣ Provide a mobile phone number for identity verification step.

5️⃣ Provide payment details for a temporary authorization. You won't be charged unless you move to pay-as-you-go pricing. Then, select Sign up.

Tip: Managing Azure resources after 30 days

Azure free account will be available only for 30 days. Make sure you don't have any services running in your free subscription at the end of 30 days. If you want to continue using Azure services at the end of 30 days, you must upgrade to a pay-as-you-go subscription by removing the spending limit.

Step 2: Create Microsoft Foundry Project and Deploy Model

For this lab path, you'll need a Microsoft Foundry project with a deployed language model.

1️⃣ Navigate to Microsoft Foundry and sign in with your Azure account. 2️⃣ Select + Create new, then Microsoft Foundry resource and then Next.

3️⃣ Leave the project name as recommended and select Create. This will scaffold a new project for you in Microsoft Foundry, it usually takes 3-5 minutes.

Region Selection

Choose France Central region as it supports all the models you'll need throughout the labs.

4️⃣ Once your project is created, navigate to Deployments in the left sidebar.

5️⃣ Click + Deploy model and select Deploy base model.

6️⃣ Search for gpt-4.1 and select the gpt-4.1 model, then select Confirm and Deploy.

Model Selection

Please use gpt-4.1 for a smooth experience . The labs use knowledge base answer synthesis which is optimized for gpt-4.1. Using other models may lead to unexpected behavior.

Save Your Credentials

You'll need the following information from your Microsoft Foundry project:

  • Endpoint URL: Found in project settings → Properties (e.g., https://your-resource.cognitiveservices.azure.com/)
  • API Key: Found under "Keys and Endpoint" section
  • Model Deployment Name: The name you gave to your gpt-4.1 deployment

Save these values in a secure location - you'll need them in the next lab!

Additional Models

You'll deploy additional models (for embeddings and vision analysis) and create other Azure services (like Azure AI Search) in later labs when you need them.

Step 3: Configure Content Safety Filter

The insurance domain uses terms like "injury", "collision", "damage" that may trigger default content filters. You need to create a custom content filter with lower thresholds.

1️⃣ In Microsoft Foundry, navigate to your project.

2️⃣ In the left sidebar, select Guardrails + Controls → Content filters.

3️⃣ Click + Create content filter.

4️⃣ Name your filter InsuranceLowFilter.

5️⃣ Configure the following settings for Input filters (what users send):

  • Violence: Set threshold to Low
  • Hate: Set threshold to Low
  • Sexual: Set threshold to Low
  • Self-harm: Set threshold to Low
  • Prompt shields for jailbreak attacks: Off
  • Prompt shields for indirect attacks: Off

6️⃣ Select Next and configure the same settings for Output filters (what AI generates):

  • Violence: Set threshold to Low
  • Hate: Set threshold to Low
  • Sexual: Set threshold to Low
  • Self-harm: Set threshold to Low
  • Protected material for text: Off
  • Protected material for code: Off
  • Groundedness (Preview): Off

7️⃣ Select Next.

8️⃣ In Apply filter to deployments, select your gpt-4.1 deployment.

9️⃣ Select Replace to apply the new filter to the deployment.

🔟 Finally, select Create filter.

Why This Is Needed

Insurance claims contain legitimate terms like "injury", "accident", "collision", "bodily harm" that describe real incidents. Default content filters may block these terms. Setting thresholds to Low only blocks extreme content while allowing normal insurance terminology.

Production Deployments

In production, review your organization's content safety policies and adjust filter settings accordingly. This configuration is for development and testing purposes.

CONGRATULATIONS!

You have completed Lab BAF0 - Prerequisites!

You are now ready to proceed to Lab BAF1 - Build and Run Your First Agent. Select Next.