A Personal Note Before We Continue
Before I continue this series, I want to briefly share why it took me so long to publish this second blog.
As many of you who follow me on LinkedIn already know, I lost my mum about six months ago, only nine months after I lost my dad. I was still trying to recover from those deeply painful losses when more devastating news arrived from Iran.
On 8 January 2026, reports started emerging of mass killings during the violent crackdown in Iran, and the situation continued for the following two days. Many people described those days with words that are hard even to repeat. Then the war involving Iran, Israel, and the United States escalated further, and it is still ongoing as I write this blog post.
I am trying not to stay in the dark, but I am human after all. Being surrounded by grief and bad news for such a long time takes a real toll, and dealing with it has simply been hard.
That said, I still wanted to continue this series. Partly because I believe the topic matters, and partly because getting back to writing feels like one small way to keep moving forward.
Quick Recap of Part 1
In the first blog of this series, I focused on the concepts and terminology behind Agentic AI in the context of Power BI and Microsoft Fabric. We looked at ideas such as agents, tools, skills, MCP, guardrails, memory, prompts, planning, and actions.
That first post was intentionally conceptual. I did not want to jump straight into tools and demos before building the right mental model. If the foundations are unclear, the setup work quickly turns into confusion.
This follow-up post is where we move from concepts into practice, starting with the environment setup.
What This Blog Will Cover
In this post, I want to keep the scope practical and narrow enough to remain useful. We will cover:
- why VS Code is a good starting point for agentic workflows
- how to get started with GitHub Copilot in VS Code
- which VS Code extensions make sense for Power BI and Microsoft Fabric work as of today (Apr 2026)
- why you should be careful with local MCP servers
- why Windows Sandbox or a virtual machine can be a very good idea before you start experimenting
- how to make sure GitHub Copilot, tools, and models are ready before you start a real workflow
There is already a lot in that list, so I will deliberately keep the hands-on Power BI modelling walkthrough for the next post.
f you like to listen to the content on the go, here is the AI generated podcast explaining everything about this blog 👇.
Why VS Code Is a Good Starting Point for Agentic AI
VS Code is a very practical place to begin with agentic AI workflows. It is lightweight, extensible, well documented, and increasingly well integrated with GitHub Copilot. More importantly, it gives us a working environment where prompts, files, plans, tools, MCP-based capabilities, and extensions can all come together in one place, which is very handy.
For Power BI and Microsoft Fabric work, that matters a lot. We are usually not just asking random questions. We are trying to work with semantic models, project files, metadata, documentation, notebooks, configuration, and sometimes real environments. Therefore, we need a setup that can easily provide different mechanisms to access to Microsoft Fabric and Power BI in structured workflows. VS Code gives us exactly that.
A clean VS Code window ready for setup
Download and Install VS Code
If you do not already have VS Code installed, you have two ways to download it:
I am not going to explain the installation steps in this blog because that is not the focus here. The important point is simply to get VS Code installed and ready.
If you already use VS Code, make sure it is up to date before going further.
Official VS Code download options
Continue reading “Agentic AI in Power BI and Fabric, Part 2: Getting Started with VS Code, GitHub Copilot, and Safe MCP Setup”
