Step-by-Step: Build Your First Agent in the NEW Azure Foundry (With 3 Powerful Tools!)

Ready to build a production-ready AI agent in the brand-new Azure AI Foundry? 🚀 Welcome back! In this hands-on tutorial, we leave the old interface behind and dive exclusively into the new Azure AI Foundry UI. If you have been struggling to find your projects or figure out how tools work in this redesigned environment, this video is your ultimate guide. I will walk you through creating a fully functional "Insurance Policies Agent" from scratch. But this isn't just a simple chatbot—we are building an agent with real skills. You will learn how to connect it to three completely different types of tools, transforming it from a basic language model into a versatile problem-solver. What you will learn in this video: 🗺️ Navigating the New UI: We start by identifying which project types (like "AI Foundry" vs. "Hub") are compatible with the new interface so you don't get lost. 🔍 Tool #1: Azure AI Search (RAG): Watch as I connect the agent to an existing Azure AI Search index filled with JSON insurance documents. You will see exactly how to select the index and how the agent uses this data to answer specific policy questions. 🌦️ Tool #2: Custom MCP Server (Real-time Data): This is where things get advanced. I show you how to connect a custom Model Context Protocol (MCP) server. We pull in a live weather tool, demonstrating how your agent can interact with external APIs and services. You will also see the new tool approval workflow in action, giving you control over what the agent executes. 📁 Tool #3: Direct File Upload (Instant Indexing): Need to give your agent knowledge fast? I demonstrate how to generate sample data using GitHub Copilot, upload those files directly into Azure Foundry, and let the platform automatically create a search index. No pre-configuration needed! 🧠 Refining Instructions with AI: I use an LLM (DeepSeek) to help craft and shorten the perfect system prompt, ensuring the agent understands all its new tools. 🌐 Sharing Your Agent: Finally, I show you the hosted preview web app. You will learn how to get a shareable link to let your colleagues test the agent without needing access to the backend. By the end of this video, you will be confident in building, testing, and sharing multi-tool agents in the latest Azure AI Foundry interface. Chapters / Timeline: 0:00 - Introduction and navigating to the new Azure AI Foundry UI. 1:30 - Creating a new agent and understanding the new naming conventions. 2:15 - Adding the first tool: Connecting to an existing Azure AI Search index. 3:45 - Testing the agent with a prompt to query insurance policies. 5:00 - Adding a second tool: Connecting a custom MCP server for weather data. 6:30 - Demonstrating the new tool approval workflow (approve once or always). 8:15 - Refining the agent's instructions using an LLM for better task management. 9:15 - Accessing and sharing the hosted preview web app with team members. 10:30 - Adding a third tool: Generating files with Copilot and uploading them. 14:00 - Final testing and observing the agent select the correct file index. 16:00 - Conclusion, recap, and call to action.