Microsoft has taken significant steps to standardize how artificial intelligence models interact with cloud data by unveiling public previews for two distinct Model Context Protocol (MCP) servers. The Azure MCP Server provides general access to various Azure resources. Concurrently, the company introduced a server focused specifically on Azure Database for PostgreSQL Flexible Server, announced via a Microsoft Community Hub blog post.
Both releases leverage the open MCP standard, aiming to simplify development and streamline AI integration by reducing the need for custom-built connectors for different data sources.
An Open Protocol For AI Context
The core challenge MCP addresses stems from the difficulty AI models face in accessing fragmented, external data needed for context-aware operations. As AI firm Anthropic, the originator of MCP, noted when launching the protocol in November 2024, “Every new data source requires its own custom implementation, making truly connected systems difficult to scale.” MCP tackles this by establishing a standard client-server architecture using HTTP. AI applications (MCP Clients) can interact with various data sources or tools exposed via dedicated MCP Servers, which offer standardized “Tools” (functions), “Resources” (data/files), and “Prompts” (templates).
Building On Earlier MCP Adoption
The new previews from Microsoft represent an expansion of existing MCP adoption. The company had previously integrated MCP into its Azure AI platform in March 2025, specifically within the Azure AI Foundry and Azure AI Agent Service. An official C# SDK for MCP, developed in collaboration with Anthropic, was released via NuGet on April 2.
MCP was also introduced into Copilot Studio, using its connector framework. This strategy aligns with efforts under Microsoft’s new CoreAI division to foster interoperability and support diverse models and tools within the Azure ecosystem, including integrations with frameworks like Microsoft’s Semantic Kernel.
Capabilities Of The Azure MCP Server
The general Azure MCP Server (Preview) enables AI agents to interact with several key Azure services. According to the Azure SDK blog announcement, supported capabilities include:
- Azure Cosmos DB: Listing accounts, listing and querying databases, managing containers/items, and executing SQL queries.
- Azure Storage: Listing accounts, managing blob containers/blobs, listing and querying Storage tables, and getting container properties/metadata.
- Azure Monitor (Log Analytics): Listing workspaces, querying logs using Kusto Query Language (KQL), listing available tables, and configuring monitoring options.
- Azure App Configuration: Listing stores, managing key-value pairs, handling labeled configurations, and locking/unlocking settings.
- Azure Resource Groups: Listing and managing resource groups.
- Azure Tools: Direct execution of Azure CLI and Azure Developer CLI (
azd) commands, supporting template operations like discovery, initialization, provisioning, and deployment.
Authentication for this server relies on Azure Identity’s DefaultAzureCredential, which automatically attempts to use credentials from the environment, shared token cache, Visual Studio, Azure CLI, Azure PowerShell, or azd before falling back to interactive browser login. Production credentials like Managed Identity can be enabled via the AZURE_MCP_INCLUDE_PRODUCTION_CREDENTIALS=true environment variable.
Specialized Tools For PostgreSQL Databases
In parallel, the Azure Database for PostgreSQL MCP Server (Preview) provides a focused set of tools for database interaction, detailed in the Community Hub blog post:
- Listing all databases and tables (with schema) within a Flexible Server instance.
- Executing read queries to retrieve specific data.
- Performing data modifications like inserting or updating records.
- Managing table structures by creating new tables or dropping existing ones.
- Accessing server configuration details (version, compute, storage) and specific parameters when using Microsoft Entra ID authentication (the recommended method).
This server allows AI clients like Anthropic’s Claude Desktop and Microsoft’s Visual Studio Code (using GitHub Copilot Agent Mode) to interact with PostgreSQL data using natural language prompts translated into database operations.
Developer Experience And Ecosystem
Developers can access the code and setup instructions for both servers via GitHub. The general Azure MCP Server is located at the Azure/azure-mcp repository and installed using Node.js via the command npx -y @azure/mcp@latest server start. The PostgreSQL-specific server resides in the Azure-Samples/azure-postgresql-mcp repository and requires setting up a Python 3.10+ environment with specific libraries (mcp[cli], psycopg[binary], Azure SDK components).
Both repositories provide configuration examples for client tools and troubleshooting guides. While the server components themselves are free during preview, associated Azure service costs still apply. The broader MCP ecosystem, including reference server implementations and community contributions, continues to evolve around the open-source protocol provided by Anthropic.
Last Updated on April 20, 2025 7:04 pm CEST
