Microsoft Agent Framework - Copilot Studio Sample

A Python sample showing how to connect to Microsoft Copilot Studio agents using the Agent Framework. Includes both basic and advanced integration examples.

Made by iLoveAgent.ai

Quick Start

# Install dependencies
uv sync

# Set up Azure app (first time only)
./setup_azure_app.sh

# Run it
uv run main.py

What You Need

• Python 3.10 or higher
• uv package manager
• Azure subscription with Copilot Studio access
• Azure CLI (for the setup script)

Configuration

Create a .env file from the example:

cp .env.example .env

Fill in your values:

• COPILOTSTUDIOAGENT__ENVIRONMENTID - Your environment ID (format: Default-<guid>)
• COPILOTSTUDIOAGENT__SCHEMANAME - Your agent's schema name
• COPILOTSTUDIOAGENT__AGENTAPPID - Azure app client ID (set by setup script)
• COPILOTSTUDIOAGENT__TENANTID - Your tenant ID (set by setup script)

Find your environment ID in the Power Platform Admin Center.

Examples

Basic Usage (main.py)

Simple integration using environment variables:

from agent_framework.microsoft import CopilotStudioAgent

agent = CopilotStudioAgent()

# Non-streaming
result = await agent.run("What is AI?")

# Streaming
async for chunk in agent.run_stream("Tell me about Azure"):
    if chunk.text:
        print(chunk.text, end="", flush=True)

Advanced Usage (copilotstudio_explicit.py)

Explicit configuration for more control:

from agent_framework.microsoft import CopilotStudioAgent, acquire_token
from microsoft_agents.copilotstudio.client import ConnectionSettings, CopilotClient, PowerPlatformCloud

token = acquire_token(client_id="...", tenant_id="...")

settings = ConnectionSettings(
    environment_id="Default-...",
    agent_identifier="...",
    cloud=PowerPlatformCloud.PROD
)

client = CopilotClient(settings=settings, token=token)
agent = CopilotStudioAgent(client=client)

Azure Setup

Automated

The setup script handles everything for you:

./setup_azure_app.sh

Or with a custom app name:

AZURE_APP_NAME="my-copilot-app" ./setup_azure_app.sh

Manual

If you need to do it manually:

# 1. Create app
az ad app create \
  --display-name "copilot-studio-agent" \
  --sign-in-audience AzureADMyOrg \
  --public-client-redirect-uris "http://localhost"

# 2. Enable public client flows
az ad app update --id <APP_ID> --set isFallbackPublicClient=true

# 3. Register Power Platform API
az ad sp create --id 8578e004-a5c6-46e7-913e-12f58912df43

# 4. Add permission
az ad app permission add --id <APP_ID> \
  --api 8578e004-a5c6-46e7-913e-12f58912df43 \
  --api-permissions 204440d3-c1d0-4826-b570-99eb6f5e2aeb=Scope

# 5. Grant consent
az ad app permission admin-consent --id <APP_ID>

Troubleshooting

Authentication fails?

rm -f .msal_token_cache.json
uv run main.py

Can't connect to environment?

• Check your environment ID format: Default-<guid> (capital D, with hyphen)
• Make sure your agent is published in Copilot Studio
• Verify you have access in Power Platform Admin Center

Permission errors?

# Create the Power Platform service principal if missing
az ad sp create --id 8578e004-a5c6-46e7-913e-12f58912df43

# Re-grant consent
az ad app permission admin-consent --id <YOUR_APP_ID>

What's Included

• main.py - Basic example with streaming and non-streaming
• copilotstudio_explicit.py - Advanced configuration example
• setup_azure_app.sh - Automated Azure AD setup
• .env.example - Configuration template

Dependencies

• agent-framework-copilotstudio (latest preview: 1.0.0b251120)
• python-dotenv
• aiohttp

Links

• Microsoft Agent Framework
• Copilot Studio Docs
• Power Platform Admin

About

This sample is maintained by iLoveAgent.ai. We help teams build and deploy AI agents with Microsoft Copilot Studio and Agent Framework.

Need help with your project? Check out iLoveAgent.ai.

License

MIT License - see LICENSE file for details.