Copilot Studio's tool surface has grown significantly, and most people are still treating it like it only has connectors. This session is a comprehensive look at every tool type available to your agent today: connectors, REST APIs, prompt actions, Power Automate agent flows, and the newest addition, Model Context Protocol (MCP). Each one unlocks different capabilities, and knowing when to reach for which one is the difference between an agent that feels powerful and one that feels patched together.

We'll start with how generative AI uses tools at runtime: how the orchestrator decides which tool to call, what happens when it picks the wrong one, and how your tool descriptions and input schemas directly influence that decision. Most developers skip this part and then wonder why their agent ignores a perfectly functional action.

From there we'll work through each tool type with live demos. By the end you'll understand not just how to add tools, but how to design them so your agent's AI layer uses them correctly, consistently, and without surprises in production.