Introduction

For the past 12 years, my journey with Power BI has been nothing short of fascinating. Every few years, Microsoft introduces something that reshapes how we interact with data and solve business problems. I still remember the early days when features like Bookmarks, Visual Interactions, and Sync Slicers, Custom Visuals completely changed the way we designed report navigation and storytelling. Back then, these capabilities felt revolutionary, opening the door for more dynamic and interactive dashboards.

Then came other powerful capabilities:

• Cortana Voice Interaction – enabling voice-driven analytics (Inactive Now)
• SharePoint & Web embedding – bringing analytics everywhere
• Power BI Apps – simplifying enterprise distribution
• Paginated Reports – enabling pixel-perfect, print-ready reports for operational and regulatory needs
• Power Platform Integration (Power Apps & Power Automate) – enabling users to trigger workflows, write back data, and build applications directly from Power BI reports
• Deployment Pipelines – improving governance and DevOps for BI
• Microsoft Fabric – unifying data engineering, data warehousing, real-time analytics, and BI into a single integrated platform
• And recently… Copilot, bringing AI-assisted analytics into the platform

Each of these features addressed a real business need and kept Power BI aligned with the evolving technology landscape.

Now we are in the AI era.

Once again, Microsoft has surprised the community with something powerful: Model Context Protocol (MCP).

When Microsoft released the MCP Server about 4 months ago, I had the same question many others had:

👉 What exactly can this do?

After spending time:

• Setting up the MCP server environment on Windows
• Exploring its capabilities
• Experimenting with Claude AI as a chat interface
• Testing how prompts interact with the Power BI semantic model

I can confidently say the potential is much bigger than what I initially imagined.

Yes, the setup may take some time. But once it works, the outcome is truly rewarding.

My initial assumption

At first, I thought MCP would mainly help end users interact with data through chatbots.

But after exploring it deeper, I realized something even more valuable:

💡 MCP is a powerful tool for developers and designers as well.

It enables you to:

• Interact directly with the semantic model
• Perform on-the-go development
• Generate and optimize DAX measures
• Improve model structure using natural language prompts
• Analyze model metadata
• Automate modeling tasks

While Copilot introduced AI-assisted development earlier, the experience was still somewhat limited.

With MCP, you now have much more freedom and control.

You can literally:

• Ask the model to generate measures
• Optimize model structure
• Improve naming conventions
• Analyze relationships and metadata
• And even converse with your data in simple human language.

Two powerful capabilities of MCP

1️⃣ Modeling MCP Server Build and modify semantic models programmatically.

2️⃣ Remote MCP Server Query data and generate insights from existing models.

With MCP, you can do both development and analytics interaction.

So what’s the real takeaway?

Technology will keep evolving.

Features will come and go. AI will definitely reduce manual workload.

But one thing will always remain at the center of everything:

❤️ Your Semantic Model

No matter how powerful AI becomes, the quality of insights will always depend on:

• How well your data is modeled
• How clean your relationships are
• How meaningful your measures and logic are
• And the essence of a good design/UI, because the way insights are presented

Editor’s Note:This article was originally published on my LinkedIn profile. I’m republishing it as part of the AI-Driven BI Insights newsletter so future editions are available in one place for subscribers.

So the biggest takeaway from exploring MCP is simple:

👉 Love your data. 👉 Invest time in your semantic model. 👉 Understand your business logic deeply.

Because when your model is strong, AI becomes a true accelerator.

Exciting times ahead for the Power BI community!

Want to try this yourself?

You only need a few things:

Your Windows machine + Visual Studio Code + a GitHub account + Claude Desktop +

Your Power BI file containing the semantic model

That’s it.

Just a few configurations and a bunch of prompts related to your data…

…and here you go!

About the author

M, Rahmatullah (25/08/2026) (2) Edition #1 – Microsoft MCP + Claude AI: A New Era of Interacting with Data | LinkedIn