---

name: agent-framework description: Build agentic AI solutions using Microsoft Agent Framework (Python). Use when creating AI agents, multi-agent workflows, chatbots, or assistants that need tool calling, streaming, orchestration patterns (sequential, concurrent, handoff, group chat), or deployment to Azure Container Apps. Covers single agents, multi-agent systems, Azure OpenAI integration, and containerized deployment.

Microsoft Agent Framework Skill

Build production-grade agentic AI solutions using Microsoft Agent Framework with Python.

Prerequisites

• Python 3.10+
• Azure OpenAI service endpoint with deployed model
• Azure CLI authenticated (az login)
• User has Cognitive Services OpenAI User or Cognitive Services OpenAI Contributor role

Installation

Copypip install agent-framework --pre

Environment Variables

Copyexport AZURE_OPENAI_ENDPOINT="https://<resource>.openai.azure.com"
export AZURE_OPENAI_DEPLOYMENT_NAME="gpt-4o"

Quick Start: Basic Agent

Copyimport asyncio
from agent_framework.azure import AzureOpenAIChatClient
from azure.identity import AzureCliCredential

agent = AzureOpenAIChatClient(credential=AzureCliCredential()).create_agent(
    instructions="You are a helpful assistant.",
    name="Assistant"
)

async def main():
    result = await agent.run("Hello, how can you help me?")
    print(result.text)

asyncio.run(main())

Quick Start: Streaming Response

Copyasync def main():
    async for update in agent.run_stream("Tell me a story."):
        if update.text:
            print(update.text, end="", flush=True)
    print()

asyncio.run(main())

Quick Start: Multimodal Input

Copyfrom agent_framework import ChatMessage, TextContent, UriContent, Role

message = ChatMessage(
    role=Role.USER,
    contents=[
        TextContent(text="Describe this image."),
        UriContent(uri="https://example.com/image.jpg", media_type="image/jpeg")
    ]
)

result = await agent.run(message)

Agent Patterns

1. Single Agent with Tools

Use for agents that need external capabilities. See references/tools.md for tool patterns.

2. Multi-Agent Workflows

For complex tasks requiring specialized agents. See references/orchestrations.md for patterns:

3. Azure AI Foundry Agent

Use for managed agents with built-in tools (Code Interpreter, File Search). See references/azure-foundry.md.

Container Deployment

Deploy agents to Azure Container Apps. See references/container-deployment.md for:

• Dockerfile configuration
• requirements.txt setup
• Health check endpoints
• Environment configuration
• Azure Container Apps deployment

Example Scripts

Execute without loading into context when possible:

Code Generation Guidelines

1. Always use async/await - Agent Framework is async-first
2. Use AzureCliCredential - For local dev authentication
3. Stream for UX - Use run_stream() for interactive experiences
4. Handle errors - Wrap in try/except for production code
5. Type hints - Include for maintainability
6. Environment variables - Never hardcode credentials

Project Structure

Copymy-agent-app/
├── src/
│   ├── __init__.py
│   ├── main.py           # Entry point
│   ├── agents/
│   │   ├── __init__.py
│   │   └── assistant.py  # Agent definitions
│   └── tools/
│       ├── __init__.py
│       └── functions.py  # Custom tools
├── Dockerfile
├── requirements.txt
├── .env.example
└── azure.yaml            # Azure deployment config