Azure MCP Server: Connect AI Coding Agents to Azure Infrastructure

What Is the Azure MCP Server?

Microsoft publishes two official MCP (Model Context Protocol) servers that connect AI coding agents to Azure. The Azure MCP Server handles infrastructure: Storage, Cosmos DB, Monitor, Kubernetes, Container Apps, and deployment. The Azure DevOps MCP Server handles project management: work items, pipelines, repos, wikis, and test plans.

Both servers run on your local machine. Your Azure credentials stay local, authenticated through the same az login you already use. The MCP protocol translates natural language requests from your AI agent into Azure API calls, with access controlled by your existing Azure RBAC permissions.

Supported Azure Services

The Azure MCP Server exposes tools for interacting with Azure services through your AI agent. Each tool category maps to a specific Azure service, and all operations go through your existing Azure RBAC permissions.

The tool set continues to grow. Microsoft merges the Azure MCP Server into the broader microsoft/mcp repository on GitHub, and new service integrations ship regularly. The server auto-updates when you use npx -y @azure/mcp@latest.

Installation and Authentication

Prerequisites

• Node.js 20+ (required for both MCP servers)
• Azure CLI installed and authenticated (az login)
• Azure subscription with appropriate RBAC permissions

Install the Azure MCP Server

You can run the server on demand with npx or install it globally for faster startup.

npx -y @azure/mcp@latest server start

npm install -g @azure/mcp@latest
azure-mcp-server start

Install the Azure DevOps MCP Server

npx -y @azure-devops/mcp your-org-name

Authentication

Both servers use DefaultAzureCredential from the Azure Identity SDK. This means they automatically discover credentials from whichever method you have configured, tried in this order:

1. Environment variables (service principal credentials)
2. Managed identity (when running in Azure)
3. Azure CLI (az login, most common for local development)
4. Visual Studio / VS Code (signed-in accounts)
5. Interactive browser (fallback)

For most developers, az login is all you need. The server picks up your active session and uses your account's RBAC permissions to authorize every operation.

Configuration for AI Clients

Claude Code

Add the Azure MCP Server to Claude Code with a single command. This registers it as a local MCP server available in agent mode.

# Azure infrastructure server
claude mcp add azure-mcp -- npx -y @azure/mcp@latest server start

# Azure DevOps server (replace with your org name)
claude mcp add azure-devops -- npx -y @azure-devops/mcp contoso

Cursor

Add the server in Cursor Settings > Tools & Integrations > New MCP Server, or edit ~/.cursor/mcp.json directly for global availability.

{
  "mcpServers": {
    "azure-mcp": {
      "command": "npx",
      "args": ["-y", "@azure/mcp@latest", "server", "start"]
    },
    "azure-devops": {
      "command": "npx",
      "args": ["-y", "@azure-devops/mcp", "contoso"]
    }
  }
}

VS Code with GitHub Copilot

Create a .vscode/mcp.json file in your project root. VS Code supports input prompts for dynamic values like organization names.

{
  "inputs": [
    {
      "id": "ado_org",
      "type": "promptString",
      "description": "Azure DevOps organization name"
    }
  ],
  "servers": {
    "azure-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@azure/mcp@latest", "server", "start"]
    },
    "azure-devops": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@azure-devops/mcp", "${input:ado_org}"]
    }
  }
}

Visual Studio 2026

The Azure MCP Server is built into Visual Studio 2026 with no separate installation required. It ships as a generally available feature with native integration into Copilot agent mode.

Azure DevOps MCP Server

The Azure DevOps MCP Server is a separate package from Microsoft that gives AI agents access to your Azure DevOps organization. It covers nine domains and runs locally, so project data stays on your machine.

Cloud MCP Server Comparison

All three major cloud providers now ship official MCP servers. The breadth of coverage varies significantly.

AWS has the broadest coverage with 60+ servers spanning its entire product catalog. Azure has deeper integration with developer tooling through Visual Studio 2026 and Azure DevOps. GCP has the smallest server count but includes granular read-only/read-write controls and comprehensive logging.

For teams running Azure DevOps for project management, the Azure DevOps MCP Server is the clear differentiator. No other cloud provider ships an official MCP server for work items, pipelines, and test management.

Use Cases

WarpGrep + Azure MCP: Code and Infrastructure in One Agent

The Azure MCP Server gives your agent access to infrastructure. WarpGrep gives it access to your codebase. Running both MCP servers together creates an agent that understands your code and the cloud environment it runs in.

Practical example: you ask "why is the checkout endpoint slow?" WarpGrep finds the relevant handler code and its database queries. The Azure MCP Server queries Log Analytics for latency metrics on that endpoint. The agent correlates slow Cosmos DB queries in the code with actual performance data from production, and suggests both a code fix and a throughput adjustment.

This is the compound effect of MCP servers. Each server adds a capability. Together they give the agent enough context to reason across layers of your stack, from source code to infrastructure metrics.

Frequently Asked Questions

What is the Azure MCP Server?

Microsoft's official implementation of the Model Context Protocol for Azure services. It lets AI coding agents interact with Azure Storage, Cosmos DB, Monitor, Kubernetes, Container Apps, and deployment tooling through natural language. It runs locally, authenticates via Azure CLI or managed identity, and works with Claude Code, Cursor, VS Code, and Visual Studio 2026.

How do I install the Azure MCP Server?

Run npx -y @azure/mcp@latest server start for on-demand usage, or npm install -g @azure/mcp@latest for a global install. You need Node.js 20+ and must be signed into Azure with az login.

What is the difference between Azure MCP Server and Azure DevOps MCP Server?

The Azure MCP Server connects to Azure cloud infrastructure (Storage, Cosmos DB, Monitor, AKS). The Azure DevOps MCP Server connects to Azure DevOps project management (work items, pipelines, repositories, wikis, test plans). They are separate npm packages that can run simultaneously in the same AI client.

Does the Azure MCP Server work with Claude Code?

Yes. Run claude mcp add azure-mcp -- npx -y @azure/mcp@latest server start to register it. For Azure DevOps, use claude mcp add azure-devops -- npx -y @azure-devops/mcp your-org-name. Both work in Claude Code's agent mode once registered.

Is the Azure MCP Server secure?

The server runs locally and never sends credentials to external services. Every operation respects your Azure RBAC permissions. All API calls are logged to an audit resource. Credentials are managed through the official Azure Identity SDK, not stored or cached by the MCP server itself.

Can I use a service principal instead of Azure CLI?

Yes. Set the AZURE_CLIENT_ID, AZURE_TENANT_ID, and AZURE_CLIENT_SECRET environment variables. The DefaultAzureCredential chain picks these up automatically. This is the recommended approach for CI/CD pipelines and shared environments.

Which Azure services are supported?

As of March 2026: Blob Storage, Cosmos DB, Monitor (Log Analytics), Azure Kubernetes Service, Container Apps, App Configuration, deployment tooling (azd), and general Azure CLI passthrough. The tool set expands with each release.

Related Pages

• Agent Client Protocol (ACP): Open Standard for AI Agents
• Agentic Coding Tools Landscape (2026)
• Claude Code vs Cursor
• WarpGrep: Semantic Code Search MCP Server