Your Entire Agentic AI Workflow Now Inside VS Code

If you're building AI apps or agents and want everything in one place, models, prompts, evaluation, and deployments right inside your editor, this video is for you. Today, we get started with Foundry toolkit for Visual Studio Code and showcase how you can go from idea to a working AI application without leaving Visual Studio Code. Microsoft Foundry is the unified AI platform for building, grounding, evaluating, deploying, and governing AI applications and agents at scale. Available through a web-based portal as well as an SDK that integrate into the developer workflows. Through the Foundry toolkit extension for Visual Studio Code and Microsoft Foundry scale integration with GitHub Copilot, developers can seamlessly access Foundry models, agents, evaluations, and deployment workflows directly from the coding environment. They can also leverage GitHub Copilot as an AI-powered assistant to manage Foundry resources and apps without leaving Visual Studio Code. Foundry toolkit is a Visual Studio Code extension designed to help developers build, test, and deploy AI applications using generative AI models. It brings your entire AI workflow into one place, from model discovery and experimentation to agent development and deployment. With Foundry toolkit, you can explore models using the model catalog with access to models from Microsoft Foundry, OpenAI, Hugging Face, and more. You can test prompts interactively in the playground. You can build AI agent via a low-code approach or a first-code approach. You can debug, visualize, and iterate on agents using the agent inspector. You can evaluate models with building metrics. You can customize models through fine-tuning. And you can deploy solutions to production and monitor performance, all from Visual Studio Code. So, instead of switching between multiple tools, everything happens in one integrated experience. Let's get started in Visual Studio Code. Let's start with the extension spanner, and let's search for Foundry Toolkit. Then, we click on the extension name, and we click on install. And we wait for a few seconds for the extensions to be installed. Once it's installed, we can find Foundry Toolkit among the list of installed extensions, and this is the Foundry Toolkit extension icon. So, you get familiar with it. So, at this point, we are ready to start building. But, let's take a look at the structure of the extension. We have three sections. The first section is the My Resources sections that contain the resources available to you. So, here you have your recent agents, so you can access the agents you recently built with Foundry Toolkit very easily. You have local resources, so local models, agents, and tools available on your local machine. You have Microsoft Foundry resources, which are basically resources you import from the cloud from Azure. And you have connected resources, which are resources from external providers, such as for example, GitHub hosted models. The second section is the developer tools. And this section include the tools used to build and deploy AI applications. In the discover subsection, you have the model catalog, where you can explore models from multiple providers, such as OpenAI, Anthropic, Google, Microsoft, Ollama, GitHub, etc. You have the tool catalog, where you can browse and manage tools, or even create your toolbox, so a centralized collection of tools exposed by via MCP. Then, under the build section, you have the create agent feature that allows you to create an agent via low-code or pro-code approach. You have the agent inspector that enables you to debug and visualize agent behavior. You can deploy your local agent to Microsoft Foundry using the Foundry toolkit UI here. You can create and manage your hosted agent and test them in these hosted agent playground. You can test your models and iterate over prompts in the model playground. You can convert, optimize, and evaluate models locally, and you can run fine-tuning jobs using custom dataset locally or in the cloud. The last subsection is about telemetry. So, you can analyze runtime behavior and performance with tracing. You can compare outputs against ground truth using evaluation metrics, and you can use model profiling to diagnose CPU, GPU, and NPU usage and analyze Windows ML events. Finally, you have the feedback section where you can access documentation, get support, and access feedback channels. From here, at any point, you can also use GitHub Copilot to assist you throughout your development workflow. For example, here, we get a suggested prompt to quick start with the Foundry toolkit. This uses the Microsoft Foundry toolkit skill to guide you through the agent development workflow by grounding the responses on validated data sources. And you don't need to install anything else in addition to the Foundry toolkit extension to get the Foundry skill in GitHub Copilot. So, once you get the extension, you also get the Foundry skill in GitHub Copilot. Now, this process will ask for some follow-up questions. And the first one is, "Where are you starting from?" So, let's imagine that I need to set up my Microsoft Foundry environment, so I need to create a Microsoft Foundry project in my Azure subscription. Then I get a second question about, "What do you have already?" And I have an Azure account already. So, I would click on I have an Azure subscription, but I need to create a Foundry project inside my existing setup. And so, GitHub Copilot now will leverage the Foundry toolkit skill to open the Foundry project creation command. So now, it's asking for approval for it to use the create project command, so I'll give it my approval. And this will require to sign in to Azure. This will open a window to continue my login process, so I select my Azure account I want to connect to. Now, it's asking for my subscription name. So, I select the subscription I want to use. Then it prompts to choose a resource group among the list of available resource group, but I want to create a new one. And I'll enter a new resource group name, FDK project. And the region, uh I'm going to go with Sweden Central. And I'm going to select uh as project name FDK demo. And I wait a few seconds for the project to be created with all the resources, which are basically two new resources in this new resource group. Uh the Microsoft Foundry resource and the Microsoft Foundry project. Okay, so my project has been deployed successfully. So now, if I go back to my resources, yes, I should be able to see my new project connected to my my extension. And if I expand this section, I can see and manage all the resources within my project, which is basically my workspace in Foundry that can that I can access from within the Foundry toolkit Visual Studio Code. So, I can inspect models, which are deployed models, and endpoint details in my Foundry project. Of course, I don't have any models yet. I just set up my project. I can look at prompt agents. So, if I create any prompt agent from within the Foundry toolkit extension, or if I create any prompt agent in the same project in the portal, I would be able to see them from here. Then, I can see and manage workflows, which are declarative workflows that orchestrate agents and business logic. I can manage hosted agents, tools, things like vector stores and data stores used by my agents, classic kind of resources like agents and threads from the classic Foundry experience. That's it. You're now ready to get started with Foundry toolkit in Visual Studio Code. In the next video, we'll dive deeper into the model catalog and playground and start experimenting with models hands-on.