Step 1: Set Up Your Development Environment

I am using VS Code, so you don’t need to manually install Python on your system (unless you want to).
VS Code can handle Python via extensions which makes it super easy and everything at once place. 

You can download complete sample : here

Install These Extensions in VS Code

Open VS Code → Extensions → install:

1. Python (Microsoft)
2. Dev Tunnels (optional but helpful)
3. GitHub Pull Requests & Issues ( Saves a lot of Download time)
   
   

    

These ensure:

• You can run Python files directly inside VS Code
• IntelliSense / linting works
• Dev tunnel commands work in the integrated terminal

Step 2: Download/Clone the Official Sample

I am using this exact sample: Agent SDK Python Cards

Run these commands inside VS Code Terminal:

git clone https://github.com/microsoft/Agents.git cd Agents/samples/python/cards

You now have the complete working Python Agent sample.

Step 3: Install All Required Packages

Inside the cards folder, run:

pip install -r requirements.txt

This installs:

• FastAPI
• Uvicorn
• Agents SDK
• dotenv

VS Code will automatically detect and configure a Python interpreter for you. Once done, your requirements.txt file will look like :

 

Step 4: Add Your M365 Agent Configuration

Inside the folder, you’ll see:

.env.TEMPLATE

Rename it to:

.env

Then open the file and fill in:

CONNECTIONS__SERVICE_CONNECTION__SETTINGS__CLIENTID=
CONNECTIONS__SERVICE_CONNECTION__SETTINGS__CLIENTSECRET=
CONNECTIONS__SERVICE_CONNECTION__SETTINGS__TENANTID=

Here I am creating a single tenant bot, hence I am suing these settings for MSI it will be different
You can refer the different type of available authentication types here

Python -m src.main

Where do these values come from?
Your Azure portal -> App Registration/ Managed Identity ( Depending on what type of application is created)

Step 5: Run the M365 Agent Locally

Start your Agent:

Python -m src.main

 

You will see :

But you will not be able to test the bot here locally, so we would need additional tools to help us test locally.

Step 6: Create a Dev Tunnel

You must expose your local bot over HTTPS.
For that we use devtunnel. 

Step 6.1 — Authenticate devtunnel

You must authenticate first or you’ll get:

Unauthorized tunnel creation access

So run:

devtunnel user login

A browser pops up  -> Sign in with the same Microsoft account used for your M365 Agent.

Step 6.2 — Create the Tunnel

Now run:

devtunnel host -p 3978 --allow-anonymous

You will get a public HTTPS URL like:

Copy this URL and we can test the bot in Azure bot service

 

Step 7: Update the M365 Agent Endpoint in Portal

Go to your Azure portal → ABS Agent → Settings → Endpoint URL
Paste: <tunnel-url>/api/messages

Click Save.

At this point:

• Your Agent is running locally
• Your tunnel is publishing it
• You will be able to can talk to your Agent

Step 8: Test the Agent (The Fun Part)

Go to your Azure bot service → Test in Web Chat.

Type:

hello

You should get back the card responses from the sample.

If the sample sends Adaptive Cards or text messages, you will see them appear here exactly as coded.

 

That's It! You Built Your First Python M365 Agent

This guide took you from:
✔ VS Code setup
✔ Python environment extensions
✔ Cloning the sample
✔ Adding env configuration
✔ Running the Agent
✔ Creating a dev tunnel
✔ Testing in Web Chat

Happy Learning!