Basic RAG is table stakes. Your enterprise customers need more. In this session, we go beyond simple vector search and build a fully agentic knowledge base that intelligently selects sources, calls live MCP servers, and returns grounded, contextual answers over enterprise data at scale.

You should already be comfortable with Azure AI Search, have built at least one RAG pipeline, and understand basic concepts around embeddings and chunking strategies.

Takeaways:

Discover how Model Context Protocol (MCP) transforms static RAG into dynamic, tool-based source selection

Learn how to connect indexed enterprise content with live MCP endpoints in a single knowledge pipeline

Understand how to evaluate and measure agentic retrieval quality in production

Uncover the failure modes of agentic RAG that will burn you in production and how to avoid them

Assumed Knowledge:

You should already have built at least one RAG pipeline and be comfortable with Azure AI Search, embeddings, and chunking strategies.

Practical Takeaways:

How MCP transforms static RAG into dynamic, tool-based source selection How to connect Azure AI Search with live MCP endpoints in a single pipeline How to evaluate agentic retrieval quality and avoid common production failure modes

Out of Session Scope:

We will NOT cover governance, compliance frameworks, or Azure AI Search pricing tiers