DEV Community

Arppith Andrews
Arppith Andrews

Posted on

Ignite Is Tonight — Here's How to Actually Build Your First AI Agent

The Ignite keynote is tonight, and if you listen to nothing else this week, listen for one word: agents. Not chatbots that answer a question and stop, but agents that can plan, call tools, pull in your data, and finish a task.

Everyone will hear about agents tonight. The more useful question is how you actually build one — and there is a free Microsoft Learn path that walks you through it, end to end, with Microsoft Foundry: Develop AI Agents on Azure.

It is 9 modules, no cost, and you come out with a working agent instead of just notes. Here is what is inside, and how I would approach it.

Start with the build loop

The first module, Develop AI agents with Microsoft Foundry and Visual Studio Code, is where the path earns its name. You build, test, and deploy an agent using the Microsoft Foundry Agent Service, both through the Azure portal and with the Foundry Toolkit for Visual Studio Code. That matters more than it sounds: the difference between reading about agents and building one is having a loop where you can change something, run it, and see what the agent actually does. Get that loop working first, before you worry about anything advanced.

Then give the agent hands: tools

An agent without tools is just a chat window. Modules two and three fix that. First you integrate custom tools, so the agent can call functions you wrote for your own problem. Then you connect MCP-hosted tools, which is the part I find most interesting — MCP (Model Context Protocol) lets an agent reach tools other people built, dynamically, instead of you hard-wiring every capability yourself. If you have ever wished your project could just talk to another service, this is the module that shows you how.

Then give the agent memory: grounding

The fourth module is the one I would not skip: Build knowledge-enhanced AI agents with Foundry IQ. An agent that only knows what the model was trained on will guess confidently about your data. Foundry IQ connects the agent to enterprise knowledge with retrieval-augmented generation (RAG), so answers are grounded in your own documents and come back with citations. The module also covers tuning retrieval quality and writing instructions that keep responses consistent.

Module five adds Web IQ, which pulls current information from the web into the agent — useful when the answer depends on something that changed yesterday rather than something in your files.

Then ship it where people actually work

Module six, Integrate your agent with Microsoft 365, is the payoff: you publish your Foundry agent to Microsoft Teams and Microsoft 365 Copilot, and connect it to workplace data with Work IQ. An agent that lives in a notebook is a demo. An agent your study group or team can actually message in Teams is a tool.

Then go further: workflows and multi-agent systems

The last three modules are the stretch goals. Module seven covers agent-driven workflows in Foundry — orchestrating agents and other components into an application. Module eight introduces the Microsoft Agent Framework for building agents in code. Module nine brings it together: orchestrating a multi-agent solution, where several agents collaborate on one problem.

You do not need all nine on day one. My suggestion: do modules one through four this week, ship something small to Teams, and come back for workflows and multi-agent orchestration once you have felt the basic loop.

What would you build first?

The path suggests you come in with fundamental AI concepts and some familiarity with Azure services, so if those are new, the Get started with artificial intelligence path on Learn is the right warm-up. After that, the honest test is a first project: a study assistant grounded in your own notes with Foundry IQ, a small support agent for a club or society, or a tool-calling agent for a workflow you repeat every week.

If you build your first agent this week, what would you make it do? I am genuinely curious — drop it in the comments.

The full path is free on Microsoft Learn: Develop AI Agents on Azure.

Top comments (0)