Agentic retrieval is rapidly becoming the backbone of modern AI systems. Beyond classic RAG pipelines, today’s agentic models can plan, query, refine, and reason across heterogeneous data sources using structured tools, retrieval functions, and MCP‑based connectors. What does it mean for an LLM to “know” when to retrieve? How do models evaluate whether their current context is sufficient? And how do retrieval tools, embeddings, and protocols like MCP combine to create multi‑step, tool‑aware search behaviors that feel almost autonomous?

In this demo‑intensive session, Alan will break down the mechanics of agentic retrieval from the ground up. Starting with the evolution from classic RAG to retrieval‑augmented agents, he will explore how function calling enables dynamic search, how retrieval tools are exposed to models, and how prompting patterns influence query generation. You’ll learn how agents refine retrieval queries, chain multiple retrieval steps, and use MCP to access structured and unstructured knowledge sources across services. The session also covers practical guidelines for designing retrieval tools, shaping agent behavior, and building reliable retrieval‑driven workflows.

Join this session if you want to understand what goes on under the hood of agentic retrieval, tool‑aware search, and MCP‑powered knowledge access—and how to build retrieval systems that think, adapt, and act.