Azure AI Agent Service + Azure AI Search MCP Server

What is Azure AI Agent Service + Azure AI Search MCP Server?

A Model Context Protocol (MCP) server that enables Claude Desktop to search content using Azure AI services. It offers two implementations: Azure AI Agent Service (recommended) with document and web search, or direct Azure AI Search with keyword, vector, and hybrid search methods. This tool is designed for developers who want to integrate Azure search capabilities into Claude Desktop workflows.

How to use Azure AI Agent Service + Azure AI Search MCP Server?

Set up a Python virtual environment (Python 3.10+), install dependencies with uv pip install, and create a .env file with required Azure credentials. Then configure Claude Desktop’s JSON settings to point to either azure_ai_agent_service_server.py or azure_search_server.py, providing the necessary environment variables. Restart Claude Desktop to see the MCP tools (hammer icon) and start searching.

Key features of Azure AI Agent Service + Azure AI Search MCP Server

• AI-enhanced search results from Azure AI Agent Service
• Search both private documents and the public web
• Web search results include source citations
• Choose between Agent Service or direct Azure AI Search
• Seamless integration with Claude Desktop
• Customizable via tool instructions and new tools

Use cases of Azure AI Agent Service + Azure AI Search MCP Server

• Query indexed documents with AI-optimized results
• Perform web searches with cited sources through Bing grounding
• Run hybrid (keyword + vector) searches on Azure AI Search indices
• Combine private document knowledge with public web information

FAQ from Azure AI Agent Service + Azure AI Search MCP Server

What are the two implementations and which should I use?

Azure AI Agent Service (recommended) uses AI-enhanced search with document and web search capabilities. Direct Azure AI Search offers keyword, vector, and hybrid search. Choose Agent Service for richer results with web grounding; choose direct integration for simpler configuration and full control.

What are the runtime requirements?

Python 3.10 or higher, the latest Claude Desktop, and an Azure AI Search service with an indexed (preferably vectorized) index. For the Agent Service implementation you also need an Azure AI Project with connections to Azure AI Search and Bing.

How do I configure Claude Desktop?

Add a JSON block under mcpServers in your Claude Desktop config, specifying the Python command, script path, and environment variables for your chosen implementation (e.g., PROJECT_CONNECTION_STRING, AZURE_SEARCH_SERVICE_ENDPOINT).

How do I authenticate to Azure?

For the Agent Service, use az login to authenticate. For direct Azure AI Search, provide the service endpoint, index name, and an API key in the .env file or Claude Desktop environment variables.

Can I customize the server’s behavior?

Yes. You can modify tool instructions, add new tools using the @mcp.tool() decorator, adjust response formatting, and change web search parameters (e.g., focus on specific domains).