Build Custom AI Agents with Azure AI Foundry

Building AI Agents with Azure AI Foundry

Azure AI Foundry Agent Service allows you to build autonomous AI agents that can reason, plan, and execute multi-step tasks using tools, functions, and grounded data.

What Makes an AI Agent Different from a Chatbot?

Feature	Chatbot	AI Agent
Conversation	Responds to prompts	Plans and executes autonomously
Tools	None	Code interpreter, file search, custom functions
Memory	Session-based	Persistent threads with history
Data	Static training data	Grounded with your real-time data

Prerequisites

• Azure AI Foundry project (see the Getting Started tutorial)
• An Azure OpenAI GPT-4o deployment
• (Optional) Azure AI Search for grounding

Step 1: Create an Agent

from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential

# Connect to your project
client = AIProjectClient.from_connection_string(
    conn_str="your-connection-string",
    credential=DefaultAzureCredential()
)

# Create an agent
agent = client.agents.create_agent(
    model="gpt-4o",
    name="my-research-agent",
    instructions="""You are a helpful research agent.
    Use your tools to find information and provide accurate,
    well-cited answers."""
)

Step 2: Add Tools

Code Interpreter

Lets the agent write and execute Python code for data analysis, chart generation, and calculations.

from azure.ai.projects.models import CodeInterpreterTool

agent = client.agents.create_agent(
    model="gpt-4o",
    name="data-analyst-agent",
    instructions="Analyze data using code interpreter.",
    tools=[CodeInterpreterTool()]
)

File Search (RAG)

Connects the agent to your documents via Azure AI Search.

from azure.ai.projects.models import FileSearchTool

agent = client.agents.create_agent(
    model="gpt-4o",
    name="knowledge-agent",
    instructions="Answer questions based on the provided documents.",
    tools=[FileSearchTool(vector_store_ids=["vs_abc123"])]
)

Custom Functions

Define your own functions that the agent can call.

from azure.ai.projects.models import FunctionTool

tools = FunctionTool(functions=[{
    "name": "get_weather",
    "description": "Get current weather for a location",
    "parameters": {
        "type": "object",
        "properties": {
            "location": {"type": "string", "description": "City name"}
        },
        "required": ["location"]
    }
}])

Step 3: Create a Thread and Run

# Create a conversation thread
thread = client.agents.create_thread()

# Add a user message
client.agents.create_message(
    thread_id=thread.id,
    role="user",
    content="Analyze the Q4 sales data and create a summary chart."
)

# Run the agent
run = client.agents.create_and_process_run(
    thread_id=thread.id,
    assistant_id=agent.id
)

# Get the response
messages = client.agents.list_messages(thread_id=thread.id)
print(messages.data[0].content[0].text.value)

Step 4: Deploy to Production

1. Test thoroughly in the Foundry playground
2. Set up content filters and safety guardrails
3. Deploy as an API endpoint
4. Monitor with Azure Monitor and Application Insights

Resources

• Azure AI Agent Service Documentation
• Agent Service Quickstart
• Microsoft Learn Training

Video: Find the official walkthrough on Microsoft Developer YouTube — search for "Azure AI Agent Service"