Semantic Kernel Agent Orchestrator using MCP Tools via Azure API Management

Model Context Protocol with Azure API Management to enable plug & play of tools to LLMs

🔧 Prerequisites

• azd, used to deploy all Azure resources and assets used in this sample.
• PowerShell Core pwsh if using Windows
• Python 3.11
• An Azure Subscription with Contributor permissions
• Sign in to Azure with Azure CLI
• VS Code installed with the Jupyter notebook extension enabled

Architecture

• Model Context Protocol servers runing behind with Azure API Management to enable plug & play of tools to LLMs. The API Management can ensure end-to-end authentication and authorization, using credential manager manager for managing OAuth 2.0 tokens to backend tools and client token validation [TO BE IMPLEMENTED].

Instructions

1. Python Environment Setup
   

   python3.11 -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   pip install -r requirements.txt

2. Create the infrastructure
   This sample uses azd and a bicep template to deploy all Azure resources, including Azure AI Search.

   • Login to your Azure account: azd auth login

   • Create an environment: azd env new

   • Run azd up.

   • Choose your Azure subscription.
   • Enter a region for the resources.

   The deployment creates multiple Azure resources and runs multiple jobs. It takes several minutes to complete. The deployment is complete when you get a command line notification stating "SUCCESS: Your up workflow to provision and deploy to Azure completed."

3. Running the Notebook with the Orchestrator
   Open the notebook orchestrator-model-context-protocol and execute it to see the orchestrator in action.

4. Delete the Resources
   You can delete the infrastruture created before by using azd down --purge