Integrate MCP Tools with Azure AI Agents

 Integrate MCP Tools with Azure AI Agents

Why Tool Discovery Matters

  • As AI agents become more powerful, they need access to many tools and services

  • Manually:

    • Registering tools

    • Updating integrations

    • Managing authentication
      becomes complex and time-consuming

  • Dynamic tool discovery solves this problem


Microsoft Connector Protocol (MCP)

What is MCP?

  • MCP = Microsoft Connector Protocol

  • A standard way for AI agents to:

    • Discover tools

    • Authenticate securely

    • Use tools dynamically at runtime


Advantages of MCP for AI Agents

1. Dynamic Tool Discovery

  • AI agents automatically receive:

    • List of available tools

    • Tool descriptions and capabilities

  • No hardcoded APIs in agent code

  • “Integrate once” approach

  • Reduces maintenance when tools change


2. Interoperability Across LLMs

  • MCP works with multiple LLMs

  • You can:

    • Switch models

    • Test different LLMs
      without rewriting integrations


3. Standardized Security

  • MCP provides consistent authentication

  • No need to manage:

    • Multiple API keys

    • Different auth methods per tool

  • Easier to scale AI agent deployments securely


What is Dynamic Tool Discovery?

  • A mechanism where an AI agent:

    • Does not have hardcoded tool knowledge

    • Queries an MCP server at runtime

  • The MCP server acts as a live tool catalog


Key Benefits of Dynamic Tool Discovery

  • Tools can be:

    • Added

    • Updated

    • Removed
      without changing agent code

  • Agents always use the latest tool versions

  • Tool complexity is handled by the server, not the agent


How MCP Enables Dynamic Tool Discovery

MCP Architecture Flow

  1. MCP Server

    • Hosts tools (functions)

    • Tools are exposed using @mcp.tool decorator

  2. MCP Client

    • Connects to the MCP server

    • Dynamically fetches available tools

  3. Azure AI Agent

    • Receives tool definitions

    • Uses tools to fulfill user requests


MCP Tools

  • Tools are a primitive type in MCP

  • Represent executable functionality

  • Client generates function wrappers

  • These wrappers are added to:

    • Azure AI Agent’s tool definitions


Key Benefits

  • Scalability
    Add or update tools without redeploying agents

  • Modularity
    Agents focus on decision-making, not tool logic

  • Maintainability
    Centralized tool management reduces errors

  • Flexibility
    Supports multiple tools, APIs, and complex workflows


When MCP is Especially Useful

  • Rapidly evolving tools and APIs

  • Multiple teams managing different services

  • Agents that need to:

    • Adapt in real time

    • Interact securely with external systems

    • Perform actions beyond text generation


Summary

  • MCP enables dynamic, secure, scalable tool usage

  • Agents become adaptive and future-proof

  • Tool management shifts from agent code to a central MCP service


Integrate Agent Tools Using an MCP Server and Client

Purpose

  • To dynamically connect tools to an Azure AI Agent

  • Uses Model Context Protocol (MCP)

  • Requires:

    • MCP Server → hosts tools

    • MCP Client → fetches and registers tools for the agent

MCP Server:

Role

  • Acts as a tool registry/catalog

  • Stores all the tools the agent can use

Initialization

FastMCP("server-name")

Key Features

  • Uses:

    • Python type hints

    • Docstrings

  • Automatically generates tool definitions

  • Tool definitions are:

    • Served over HTTP

    • Centralized on the server

Benefits

  • Tools can be:

    • Added

    • Updated

    • Removed
      without modifying or redeploying the agent

  • Clean separation between tools and agent logic


MCP Client:

Role

  • Acts as a bridge between:

    • MCP Server

    • Azure AI Agent Service

Responsibilities

After initializing a session, the MCP client:

  1. Discovers tools

    • Uses session.list_tools()

  2. Generates Python function stubs

    • Wraps each tool

  3. Registers tools with the agent

    • Makes them usable at runtime


Registering Tools with an Azure AI Agent

Tool Invocation

  • MCP tools are called using:

session.call_tool(tool_name, tool_args)

Important Requirements

  • Each tool must be wrapped in an async function

  • Enables the agent to:

    • Invoke tools during conversation

    • Use them like native functions


Making Tools Available to the Agent

Process

  • Wrapped async functions are bundled into a FunctionTool

  • The FunctionTool is:

    • Registered with the agent’s toolset

  • Tools are available at runtime for user requests


MCP Agent Tool Integration Flow

  1. MCP server hosts tools using @mcp.tool

  2. MCP client connects to the MCP server

  3. The client fetches tools using session.list_tools()

  4. Each tool is wrapped in an async function

  5. Async functions call tools via session.call_tool

  6. Tool functions are bundled into FunctionTool

  7. FunctionTool is registered with the agent

  8. Agent invokes tools using natural language


Key Benefits of This Architecture

  • Dynamic – No hardcoded tools

  • Scalable – Add tools without redeploying agents

  • Modular – Tool logic separate from agent logic

  • Maintainable – Centralized tool management

  • Flexible – Agent adapts as new tools become available


Summary

By integrating an MCP server and client:

  • Tool management becomes centralized

  • Agents become adaptive and extensible

  • Systems evolve easily as new tools are introduced


Use Azure AI Agents with MCP Servers

Purpose

  • Enhance Microsoft Foundry agents by connecting them to MCP servers

  • MCP servers provide:

    • External tools

    • Contextual data

  • Extends agent capabilities beyond built-in functions


Key Advantage

  • Azure AI Agent Service supports remote MCP servers

  • Agents can quickly connect to MCP servers and use tools

  • No need to manually manage MCP client sessions


How MCP Integration Works (Simplified)

When using Microsoft Foundry Agent Service:

  • You do NOT need to:

    • Create an MCP client session

    • Manually wrap or register function tools

  • Instead:

    • Create an MCP Tool object

    • Attach the MCP server details during the agent run

This allows using multiple MCP servers in a single agent, based on need


Integrating Remote MCP Servers

Requirements

  1. Remote MCP server endpoint
    Example:

    https://api.githubcopilot.com/mcp/

  2. Microsoft Foundry agent

    • Configured to use MCP tools


Connecting Multiple MCP Servers

  • Add each MCP server as a separate tool

  • Each MCP tool includes:

PropertyDescription
    server_label        Unique identifier (e.g., GitHub)
    server_url        MCP server URL
    allowed_tools (optional)    Tools the agent is allowed to access


MCP Tool Headers

Custom Headers Support

  • MCP tools can include headers for:

    • Authentication (API keys, OAuth tokens)

    • Other required server headers

Important Notes

  • Headers are:

    • Included in tool_resources during each run

    • Not stored between runs


Invoking MCP Tools Automatically

Key Difference

  • No need to:

    • Wrap tools

    • Call session.call_tool

  • Tools are invoked automatically when needed


Steps to Use MCP Tools in an Agent Run

  1. Create an McpTool object with:

    • server_label

    • server_url

  2. Apply headers using:

    • update_headers

  3. Set approval mode using:

    • set_approval_mode

Approval Modes

  • always (default)
    Developer approval is required for every tool call

  • never
    No approval required

  1. Create a ToolSet

  2. Add the McpTool to the ToolSet

  3. Create an agent run

  4. Specify the toolset property

Handling Tool Approval (requires_action)

  • If approval is required:

    • Run status becomes requires_action

  • In requires_action, you get:

    • Tool name

    • MCP server details

    • Arguments to be passed

Approval Flow

  1. Review the tool and arguments

  2. Decide whether to approve

  3. Submit approval using:

    • call_id

    • approve = true

Results

  • When the run completes:

    • Tool results appear in the agent response

  • Agent uses tools naturally via language-based reasoning

Key Benefits of MCP with Azure AI Agents

  • Richer, context-aware agents

  • Secure external tool access

  • Multiple MCP servers per agent

  • Dynamic tool availability

  • No redeployment when tools change

Summary

  • MCP integration enables powerful, flexible, and scalable AI agents

  • As MCP ecosystem grows:

    • Agents become smarter

    • Workflows become more dynamic

    • Tool usage becomes seamless

Comments

Post a Comment

Popular posts from this blog

AI-900-3,4

Image
Module : 3 Get Started with Machine Learning in Azure Machine learning drives today’s AI by turning data into insights for predictions and recommendations. Creating an ML solution requires key decisions that affect cost, speed, and quality. Using Microsoft Azure, the process follows six main steps: Define the Problem: Decide what to predict and how to measure success. Get the Data: Find and access reliable data sources. Prepare the Data: Clean, explore, and format the data for modeling. Train the Model: Choose algorithms and adjust parameters. Integrate the Model: Deploy it to generate predictions. Monitor the Model: Track performance and retrain as needed. Machine learning is an ongoing process. Models evolve as data changes. Step 1: Define the Problem The first step in building a machine learning (ML) solution is to define the problem. You need to clearly understand what you want the model to predict and how you will measure success. Key Considerations Model output: What result shoul...
Read more

AI-900 12,13

Module : 12   Overview of Information Extraction * Information extraction is a process that uses AI to pull structured data from unstructured or semi-structured documents, like receipts, invoices, forms, and scanned files. It combines computer vision, OCR, and machine learning or generative AI. * Key Steps in an Information Extraction Pipeline 1. Text Detection & Extraction (OCR) - It uses computer vision to find text areas in an image. - The system extracts raw text from scanned or image-based documents. 2. Value Identification & Mapping It maps the OCR text to specific data fields using: - Machine learning - Rule-based mapping - Generative AI (in newer solutions) Example From a scanned receipt, the system extracts fields such as: - Vendor - Date - Subtotal - Tax - Total * Choosing the Right Approach When designing an information extraction solution, consider: Document Characteristics - Layout consistency - Fixed templates lead to simple, rule-based extraction - Many forma...
Read more

AI-900 10,11

Image
Module : 10  Introduction to Computer Vision Concepts  Computer Vision Tasks & Techniques   * What is Computer Vision? Computer vision uses AI techniques to process and understand visual input such as: - Images   - Videos   - Live camera streams   - It enables machines to "see" and interpret visual information.   1. Image Classification   - This predicts a single label for an entire image.   - The model is trained using many labeled images.   - For example, a smart grocery checkout can identify an apple, orange, or banana from a single item placed on a scale.   - Use case: Recognizing the main subject in an image.   2. Object Detection   - This identifies multiple objects in a single image.   It returns:   - Object labels   - Bounding box coordinates   - For example, a checkout system can detect all fruits placed together....
Read more