The Azure Language MCP Server

The lesson introduces the Azure Language MCP server, which wraps text analytics endpoints so LLMs can use NLP as a tool. It covers the architecture of an agent delegating to the MCP server via Foundry, the token-cost tradeoffs of letting an LLM orchestrate NLP calls versus hardcoding them deterministically, and guidance on when to use an MCP server versus performing NLP directly. It also contrasts NLP and LLM approaches to sentiment detection, noting the accuracy and cost differences, and explains how to configure the Azure Language MCP server as a tool for an agent.