Explore development approaches
Unit 3 of 10 · Module 7: Develop AI agents with Microsoft Foundry and Visual Studio Code
Read this lesson
The showroom: Foundry portal
Foundry Agent Service lets you build agents your way, from point-and-click screens to pure code, like a furniture maker with a showroom out front and a workbench in the back, so you can pick what suits your scenario and your team. The Foundry portal is a web interface for creating and managing agents without writing code. It's ideal for quick prototyping, without setting up a development environment. For visual configuration, through forms and dropdowns instead of code. For centralized management of all your agents across projects. For collaborating with stakeholders who prefer visual interfaces. And for resource oversight, with dashboards for token usage, latency, and evaluation outcomes. Nothing to install: open your Foundry project, go to Agents, and start building.
The workbench: VS Code
Foundry Toolkit for Visual Studio Code, once called the AI Toolkit, puts the same abilities right inside your code editor: that's the workbench. It has three sections. My Resources lets you browse what's in your project: models, agents, tools, and knowledge sources. Developer Tools gives you the model catalog, a model playground, Agent Builder for designing and testing agents, the Agent Inspector for debugging, and deployment for hosted agents. And Help and Feedback links to documentation and support. Agent Builder is the visual editor, it can generate the code that connects an agent to your app, and you can also edit an agent's YAML directly for precise control. VS Code suits developer-centric workflows alongside your app code, version control with Git, rapid iteration without switching tools, code-first YAML editing, and local, offline design before deploying to Azure. You install it from the Visual Studio Code Marketplace, and connect it to your existing Foundry projects.
One workflow
Whichever approach you choose, agent development follows the same pattern. Connect to your Foundry project. Create an agent with a descriptive name and purpose. Configure its instructions, to define its behavior and capabilities. Add tools to extend what it can do. Test it in the integrated playgrounds. Iterate on the design based on the results. Deploy it to production. And integrate it into your applications. The portal and the VS Code extension both support this workflow; they differ mainly in interface style, not in capability.
Resources and choices
Either way, you need the same Azure resources, like the workshop and its lumber: a Foundry project to hold your agents, models, and assets, and model deployments, like GPT-4.1 or Claude Sonnet 4.6, to power the agents. Creating a project provisions the infrastructure automatically, and as you add capabilities like file search or custom tools, the service integrates supporting services behind the scenes. Extending further, for example with Foundry IQ, may need additional Azure services. Optional services include Azure AI Search, for advanced knowledge retrieval; Azure Storage, for files your agents can access; Azure Key Vault, for secrets and credentials; and Azure Functions, for custom tools and business logic. Choose the portal for visual configuration, centralized management, or quick prototyping. Choose VS Code for developer-centric workflows, tight integration with app code, or version-controlled configuration. Plenty of teams mix them: the portal for exploring ideas and reviewing with stakeholders, VS Code for the detailed build and for shipping to production.
Facts from Microsoft's documentation: What is Microsoft Foundry Agent Service?; Microsoft Foundry Toolkit for Visual Studio Code overview (CC BY 4.0), rewritten in Loop's own words and examples. Narration, animation, and quiz by Loop. Microsoft, Azure, and Microsoft Foundry are trademarks of Microsoft. Loop is independent and not affiliated with or endorsed by Microsoft.