What AI Platforms Are Compatible with AI Agents for Beginners?

AI Agents for Beginners supports Microsoft Azure AI Foundry Agent Service V2, Azure OpenAI, OpenAI, GitHub Models, MiniMax, and any OpenAI-compatible API endpoint through its Microsoft Agent Framework (MAF) abstraction layer.

The microsoft/ai-agents-for-beginners repository is designed as a vendor-neutral learning environment for building AI agents. Its architecture centers on the Microsoft Agent Framework (MAF), which decouples agent logic from underlying model providers. This means you can write agent code once and deploy it across multiple AI platforms by changing only configuration parameters.

Supported AI Platforms and Integration Methods

Azure AI Foundry Agent Service V2

The Azure AI Foundry Agent Service V2 is the primary managed platform for deploying agents in this repository. In 02-explore-agentic-frameworks/code_samples/02-python-agent-framework.ipynb, the AzureAIProjectAgentProvider class establishes a direct connection to this service.

from agent_framework import Agent
from agent_framework.azure import AzureAIProjectAgentProvider
from azure.identity import AzureCliCredential

provider = AzureAIProjectAgentProvider(
    endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"),
    deployment_name=os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME"),
    credential=AzureCliCredential(),
)

agent = Agent(
    name="TravelRecommender",
    instructions="Help users plan trips based on preferences.",
    provider=provider,
)

The AzureAIProjectAgentProvider handles server-side agent registration, authentication via AzureCliCredential, and runtime management through the Azure portal.

Azure OpenAI and OpenAI Public API

The repository supports both Azure OpenAI and the public OpenAI API through the OpenAIChatClient class. As documented in 14-microsoft-agent-framework/README.md, this client accepts any OpenAI-compatible endpoint.

from agent_framework import Agent
from agent_framework.openai import OpenAIChatClient

client = OpenAIChatClient(
    base_url="https://api.openai.com/v1",
    api_key=os.getenv("OPENAI_API_KEY"),
    model_id="gpt-4o",
)

agent = Agent(
    name="OpenAIAssistant",
    instructions="Answer user queries.",
    chat_client=client,
)

The same OpenAIChatClient works with Azure OpenAI by substituting base_url with your Azure endpoint and using AzureKeyCredential or AzureCliCredential for authentication.

GitHub Models

GitHub Models is a Microsoft-hosted service providing OpenAI-compatible access to models including GPT-4o and Llama 3.1. The 14-microsoft-agent-framework/README.md demonstrates swapping GitHub Models into existing agent code without structural changes.

from agent_framework.openai import OpenAIChatClient

client = OpenAIChatClient(
    base_url="https://models.inference.ai.azure.com",
    api_key=os.getenv("GITHUB_TOKEN"),
    model_id="Meta-Llama-3.1-70B-Instruct",
)

This compatibility stems from GitHub Models implementing the OpenAI REST API specification, allowing the OpenAIChatClient to consume it transparently.

MiniMax and Other OpenAI-Compatible Providers

The repository explicitly supports MiniMax, a provider offering large context windows up to 204K tokens. As noted in the main README.md and 14-microsoft-agent-framework/README.md, MiniMax integrates through the same OpenAIChatClient pattern.

client = OpenAIChatClient(
    base_url="https://api.minimax.io/v1",
    api_key=os.getenv("MINIMAX_API_KEY"),
    model_id="MiniMax-M2.7",
)

This architectural pattern extends to any OpenAI-compatible endpoint, including self-hosted models with OpenAI-compatible servers like vLLM or Text Generation Inference.

Supporting Infrastructure and Services

Azure AI Search for Retrieval-Augmented Generation

While not a model provider, Azure AI Search integrates as a critical infrastructure component for RAG scenarios. Lesson 05 (05-agentic-rag/README.md) demonstrates using azure-search-documents as a vector store and retrieval tool.

from azure.search.documents import SearchClient
from agent_framework.tools import RetrievalTool

search_client = SearchClient(
    endpoint=os.getenv("AZURE_SEARCH_SERVICE_ENDPOINT"),
    index_name="knowledge-base",
    credential=AzureKeyCredential(os.getenv("AZURE_SEARCH_API_KEY")),
)

retrieval = RetrievalTool(search_client=search_client, top_k=5)
agent.add_tool(retrieval)

Agent-to-Agent (A2A) and Model Context Protocol (MCP)

The requirements.txt file includes dependencies for Agent-to-Agent (A2A) and Model Context Protocol (MCP) protocols. These emerging standards enable interoperability between agents built on different frameworks, extending compatibility beyond the core platforms listed above.

Key Configuration Files and Documentation

File	Purpose
README.md	Overview of supported platforms and setup requirements
AGENTS.md	Architectural documentation for MAF and provider abstraction
requirements.txt	SDK dependencies including azure-ai-projects, azure-ai-inference, openai
02-explore-agentic-frameworks/code_samples/02-python-agent-framework.ipynb	Azure AI Foundry Agent Service V2 implementation
14-microsoft-agent-framework/README.md	OpenAI-compatible provider configuration (OpenAI, GitHub Models, MiniMax)
05-agentic-rag/README.md	Azure AI Search integration for retrieval

Summary

• AI Agents for Beginners supports Azure AI Foundry Agent Service V2, Azure OpenAI, OpenAI, GitHub Models, MiniMax, and any OpenAI-compatible API endpoint.
• The Microsoft Agent Framework (MAF) provides provider abstraction through AzureAIProjectAgentProvider and OpenAIChatClient classes.
• Azure AI Search integrates for RAG scenarios via azure-search-documents.
• Configuration changes—not code rewrites—enable switching between platforms.

Frequently Asked Questions

How do I switch from Azure AI Foundry to OpenAI in the same project?

Update your environment variables and instantiate OpenAIChatClient instead of AzureAIProjectAgentProvider. The Agent class accepts either provider through its provider or chat_client parameter, so your agent logic remains unchanged.

Does AI Agents for Beginners support local LLMs?

Yes, any local model exposing an OpenAI-compatible REST API works with OpenAIChatClient. Configure base_url to point at your local server (e.g., http://localhost:8000/v1) and provide the appropriate model_id supported by your local deployment.

What authentication methods does Azure AI Foundry Agent Service V2 require?

The AzureAIProjectAgentProvider uses AzureCliCredential by default, falling back to DefaultAzureCredential for flexible authentication including managed identities and service principals. Your Azure subscription must have the Azure AI Foundry resource provisioned with appropriate RBAC permissions.

Are there costs associated with GitHub Models?

GitHub Models offers a free tier with rate limits for prototyping and learning. Production deployments require Azure AI Foundry or direct API provider subscriptions. Check current GitHub Models documentation for specific rate limits and pricing tiers.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}