Azure MCP Server

What is Azure MCP Server?

A Model Context Protocol (MCP) server implementation for Azure resource management and exploration. It provides a structured interface to interact with Azure resources, bridging MCP clients and Azure services via tools for managing Resource Groups and Storage Accounts. Built with Python, it supports asynchronous operations, multiple authentication methods, and structured response formats.

How to use Azure MCP Server?

Install Python 3.13+, the uv package manager, and the Azure CLI (recommended). Clone the repo, create a virtual environment, and sync dependencies from pyproject.toml. Configure environment variables in a .env file (subscription ID, tenant ID, client ID/secret). Start the server via uv run server.py for stdio transport or uv run main.py for SSE transport (listens on http://127.0.0.1:8000). Alternatively, build and run the Docker container (docker build -t mcp-azure-server:latest .). Invoke tools such as list_resource_groups with required parameters (subscription_id, auth_type).

Key features of Azure MCP Server

• Asynchronous Azure operations using modern Azure SDK
• Three authentication methods: default, service principal, managed identity
• Structured response format via MCP protocol
• Comprehensive error handling and logging
• Environment variable configuration support
• Containerized deployment option with Docker

Use cases of Azure MCP Server

• Programmatically list and inspect Azure resource groups in a subscription
• Query storage accounts and their capacity usage across a subscription
• Monitor Azure infrastructure via MCP-enabled chat assistants or custom clients
• Automate resource management workflows with service principal credentials
• Deploy in Azure-hosted environments using managed identity authentication

FAQ from Azure MCP Server

What authentication methods does the server support?

Three methods: "default" (DefaultAzureCredential for local development), "spn" (service principal using client secret), and "identity" (managed identity for Azure-hosted deployments). The auth_type parameter selects the method.

What are the runtime prerequisites?

Python 3.13 or higher, an Azure subscription, and the Azure CLI (recommended for local development). A virtual environment managed by uv is required.

How can I run the server with SSE transport instead of stdio?

Comment out the mcp.run() block in server.py and run uv run main.py. This starts a Uvicorn ASGI server on http://127.0.0.1:8000 that accepts SSE connections. Clients (e.g., MCP Inspector or VS Code Copilot) must be configured to use SSE transport.

Can I run the server inside a container?

Yes. Build the image with docker build -t mcp-azure-server:latest . then run it with environment variables (AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET). Note that DefaultAzureCredential may fail inside a container; use auth_type: "spn".

What tools are currently available?

Four tools: list_resource_groups, list_storage_accounts, list_storage_account_usage (for a specific account), and list_storage_account_usage_all (all accounts in a subscription). All require subscription_id and auth_type as input.