Microsoft has announced the Public Preview of the open-source Azure MCP Server, a new tool designed to enhance the capabilities of AI agents by providing access to Azure resources. The Azure MCP Server allows AI agents to interact with Azure services such as file storage, databases, and logs, and execute CLI commands.
Model Context Protocol (MCP) is an open protocol that standardizes the interaction between AI agents and external resources. The Azure MCP Server implements this protocol, exposing Azure services to AI systems. According to Microsoft, this enables developers to build context-aware agents for their Azure resources. For instance, agents can now query Azure Cosmos DB using natural language, access Azure Storage files, and analyze Azure Log Analytics logs.
The Public Preview of the Azure MCP Server includes support for the following Azure services and tools:
- Azure Cosmos DB (NoSQL): List accounts, databases, containers, and items; execute SQL queries.
- Azure Storage: List accounts and blob containers/blobs; manage blob containers and blobs; list and query tables; get container properties and metadata.
- Azure Monitor (Log Analytics): List workspaces and tables; query logs using Kusto Query Language (KQL); configure monitoring.
- Azure App Configuration: List stores; manage key-value pairs and labeled configurations; lock/unlock settings.
- Azure Resource Groups: List and manage resource groups.
- Azure CLI: Execute commands directly, with full functionality and JSON output.
- Azure Developer CLI (azd): Execute commands directly, supporting template discovery, initialization, provisioning, and deployment.
(Source: Medium blog post)
Brian Veldman concluded in a Medium blog post on the Azure MCP Server:
From now on, I can use the Azure MCP Server to interact with the Azure services within my subscription. This is especially helpful in troubleshooting scenarios, such as analyzing logs.
Microsoft states that this functionality allows agents to operate on Azure services, manage cloud resources, and deploy applications. Yet, more open-source projects, such as the Azure CLI MCP Server, are available on GitHub, leveraging the MCP Server for Azure resources. Julion Dubois, a principal manager, Java Developer Relations, tweets on the mentions the Azure CLI MCP Server:
It’s an MCP server that wraps the Azure CLI, so your LLM can directly send commands to Azure.
In addition, Madni Aghadi states on via tweet on X:
“MCP is just hype” That’s what I thought until I saw 1000+ MCP servers built since its launch.
Any agent that supports the MCP client pattern, including GitHub Copilot Agent Mode and custom MCP clients, can use the Azure MCP Server.
- GitHub Copilot Agent Mode: The Azure MCP Server can be installed with GitHub Copilot in VS Code. Microsoft recommends combining the Azure MCP Server with the GitHub Copilot for Azure extension for an enhanced development experience.
- Custom MCP Clients/Agents: Agents must adopt the MCP client pattern to interact with the Azure MCP Server. Frameworks like Semantic Kernel can be used to build such agents. Microsoft provides a command (npx -y @azure/mcp@latest server start) to install and execute the server, and notes that the Azure MCP Server should work with any MCP client.
The Azure MCP Server follows similar moves by other cloud providers to enhance AI agent capabilities within their ecosystems:
Lastly, Microsoft plans to enhance the Azure MCP Server with more agent samples, documentation, Microsoft products, Azure service integrations, and additional features.
