What if you could build a complete Power BI semantic model — date table, time intelligence measures, relationships and all — without touching Power BI Desktop at all?
Not a future roadmap item. Not a preview feature. Something you can set up today.
In this article, I am going to show you exactly how to do that using Claude (the AI assistant from Anthropic) and the Power BI Modeling MCP Server. By the end, you will understand what MCP is, why it matters for Power BI developers, and how to get everything connected so you can start modeling with natural language commands.
Watch the full video walkthrough here: 👉 No Code Power BI — Claude and Power BI Modeling MCP Server
What Is an MCP Server — And Why Should Power BI Developers Care?
(Watch: 0:02:51)
Before diving into the setup, let’s clear up what MCP actually means. MCP stands for Model Context Protocol, and it is the mechanism that allows AI agents and tools to take actions inside a specific environment.
Think of it this way: the AI — Claude, GitHub Copilot, ChatGPT, or any other LLM — is the brain. The MCP server is the hands. It is what allows the AI to reach into your environment and actually do things: create a table, write a DAX measure, build a relationship.
Without an MCP server, an AI tool can talk to you all day. With an MCP server connected, it can act.
For Power BI specifically, there are multiple MCP servers available today:
- Power BI Modeling MCP Server — This is what we cover in this article. It connects to a model open in Power BI Desktop (local connection).
- Remote Power BI MCP Server — This connects to your Power BI Service and semantic models published to a workspace. We will cover that in a separate article.
Today, the focus is the Modeling MCP Server: your local Power BI Desktop file, your AI tool, and the bridge between them.
What Can the Power BI Modeling MCP Server Actually Do?
The Power BI Modeling MCP Server gives an AI tool the ability to perform modeling activities on your semantic model. That includes things like:
- Creating tables (using DAX calculated tables or Power Query)
- Adding measures
- Building relationships between tables
- Organizing measures into folders
- Marking a table as a date table
- Documenting your model
What it cannot do is build reports or visuals. This is a modeling tool only. The moment Claude finishes creating your measures, you go back to Power BI Desktop to build the visuals yourself. But think about how much of your Power BI time is spent on the modeling side — that is the part this automates.
If you want to understand why data modeling is the foundation of every Power BI solution, read my article on Basics of Power BI Modeling and Star Schema.
What You Need Before Getting Started
(Watch: 0:04:00)
Here is what you need installed and configured:
- Power BI Desktop — with a semantic model open locally
- Visual Studio Code — free, open-source, download from code.visualstudio.com
- Power BI Modeling MCP Server extension — installed inside VS Code
- Claude Desktop Application — downloadable from Anthropic’s website
Let’s walk through each piece.
Step 1: Install the Power BI Modeling MCP Server Extension in VS Code
(Watch: 0:05:00)
Once you have Visual Studio Code installed and open, go to the Extensions panel (the icon on the left sidebar that looks like building blocks). Search for “PowerBI”. You will see a few results. The one you want is called “Power BI Modeling MCP Server”.
If you haven’t installed it yet, you will see an Install button. Click it.
Once installed, this extension registers the MCP server locally on your machine, and you will be able to see it listed under your installed MCP servers in VS Code.
A quick note: If you are using GitHub Copilot inside VS Code, you can actually start talking to your Power BI model directly from there after this step. But in this walkthrough, we are connecting Claude instead, because Claude is an excellent example of how any AI tool can connect to this server — the principle is exactly the same.
Step 2: Find the Path to the MCP Server
(Watch: 0:07:00)
Claude needs to know where your MCP server is installed on your machine so it can connect to it. Here is how to find that path:
- Go to your user profile folder (e.g.
C:\Users\YourName) - Navigate into
.vscode→extensions - Look for the folder that starts with
analysis-servicesorpowerbi-modeling-mcp - Go into the
serversubfolder - You will find the
powerbi-modeling-mcpexecutable file — this is what you need
Right-click on it and go to Properties to copy the full path. You will use this in the next step.
Important: When you paste this path into the Claude config file, replace every single backslash (\) with a double backslash (\\). This is a JSON formatting requirement and is one of the most common points where the setup breaks if you miss it.
Step 3: Connect Claude to the MCP Server
(Watch: 0:08:02)
Open the Claude Desktop Application. Go to:
File → Settings → Developer → Edit Config
This will open Claude’s configuration file — a JSON file — either directly or in VS Code (or any text editor).
When you first open this file, it may be nearly empty. You need to add a section that tells Claude about your MCP server. The JSON structure looks like this:
"mcpServers": {
"powerbi-modeling-mcp": {
"command": "C:\\reza\\.vscode\\extensions\\analysis-services.powerbi-modeling-mcp-0.4.0-win32-arm64\\server\\powerbi-modeling-mcp.exe",
"args": [
"--start"
],
"env": {}
}
}
Replace the path with your own — using double backslashes. Save the file. That’s it.
The next time you open Claude, it will automatically detect and connect to the Power BI Modeling MCP Server. You are now ready to model with AI.
I have put the exact config script in the description of the YouTube video so you can copy it directly: Watch here
Step 4: Connect Claude to Your Power BI Model
(Watch: 0:10:33)
Open your Power BI Desktop file. For this example, I am using a simple model with two tables: a Customer table and a FactInternetSales table — the classic AdventureWorks dataset that I use in many of my training examples.
In Claude, type something like:
“Connect to AW SM1 in Power BI Desktop”
Replace AW SM1 with the name of your own Power BI file. Claude will establish a local connection to the open model in Power BI Desktop. This is a localhost connection — it is not going through the cloud or the Power BI Service. Your model is running locally, and Claude is connecting to it directly.
Note: This local modeling connection only works with files open in Power BI Desktop. If you want to connect to a published semantic model in the Power BI Service, you need the Remote Power BI MCP Server — which I will cover in a follow-up article.
Step 5: Create a Date Table — With Natural Language
(Watch: 0:11:12)
This is where the magic starts.
With the connection established, type something like:
“Add a date table in this model starting from 2005 ending 2010, and connect it to FactInternetSales using OrderDateKey”
Claude receives this instruction. It then combines two things:
- The LLM capability — to understand your intent, figure out the best approach, and generate the appropriate DAX or Power Query code.
- The Power BI Modeling MCP Server — to actually execute that code inside your model.
In this case, Claude decided to create the date table as a DAX calculated table, and then immediately built the relationship between the date table and the FactInternetSales table.
Watch your Power BI Desktop model view as this runs. You will literally see the table appear and the relationship get created in real time — all driven by a single natural language sentence.
Once it finishes, do a quick Refresh in Power BI Desktop to load the data into the newly created table.
Want to understand the options for creating a date table in Power BI? Read my detailed comparison: Power BI Date Table: DAX or Power Query?
And if you want the full DAX script for a date table explained step by step: All-in-One Script to Create a Calendar Table Using DAX in Power BI
Step 6: Create a Measure Table With All Time Intelligence Calculations
(Watch: 0:13:37)
Now — while the model is refreshing — type the next instruction:
“Create a measure table with all time intelligence calculations for sum of sales amount from FactInternetSales table”
Claude processes this and creates:
- A dedicated measure table (blank table used as a container for measures — a best practice I always recommend)
- All the common time intelligence DAX measures, organized into folders inside the table: Year-to-Date, Quarter-to-Date, Month-to-Date, Year-over-Year, and more
In the demo, Claude created 27 measures — all properly written DAX, all organized into display folders, all based on the Sales Amount column from FactInternetSales.
This is the kind of task that would normally take an experienced Power BI developer an hour or more to write, test, and organize. Claude did it in seconds.
To understand what time intelligence calculations look like and how they work in DAX, read:
Step 7: Refresh and Visualize
(Watch: 0:14:48)
Once the measures are created, refresh your semantic model in Power BI Desktop. Now head to the report canvas and add a visualization manually.
For the demo, I dragged the Year-to-Date Sales Amount measure into a column chart and used the date hierarchy (Year, Quarter, Month) from the date table I just created. The result: a clean YTD accumulation chart that you can now use as the foundation of any report.
The AI built the model. I built the visual. That is exactly the right division of labor.
Works With PBIX and PBIP
(Watch: 0:12:35)
Whether you are using the classic PBIX format or the newer PBIP (Power BI Project) format, the Power BI Modeling MCP Server works with both. The experience is slightly different for PBIP files in terms of how changes are written, but the natural language interface remains exactly the same.
If you want to understand the difference between PBIX and PBIP, read: Power BI File Formats Explained: PBIX vs PBIT vs PBIP vs PBIR
What Comes Next
This is the local modeling scenario. Everything here runs on your machine — Power BI Desktop open, Claude connected via the MCP server, commands flowing in natural language.
The next step is the Remote Power BI MCP Server, which connects to your Power BI Service and lets you interact with published semantic models in your workspace. I will cover that in a separate video and article.
If you are using GitHub Copilot instead of Claude, the setup on the VS Code side is already done the moment you install the extension — you can start asking modeling questions directly inside the Copilot chat in VS Code. The Claude-specific steps (steps 2 and 3 above) are only needed for connecting Claude Desktop.
Summary: No Code Power BI Modeling With Claude
Here is what you can achieve once this setup is in place:
| What You Say | What Claude Does |
|---|---|
| “Add a date table from 2005 to 2010 connected to FactInternetSales on OrderDateKey” | Creates the table and relationship |
| “Create a measure table with all time intelligence calculations for Sales Amount” | Creates 27 DAX measures in display folders |
| “Mark the date table as a date table” | Applies the date table designation |
| “Add year-to-date sales amount in a column chart based on year, quarter, month” | Adds the measure (visual you do yourself) |
No DAX code written manually. No right-clicking inside Power BI Desktop. No formatting the date table from scratch. Just instructions — and a model that builds itself.
This is not replacing Power BI skills. It is accelerating them. Understanding what a date table is, why time intelligence requires one, and what YTD means — that knowledge still matters enormously, because you are the one directing the AI. The better your Power BI knowledge, the more precise and powerful your instructions will be.
Related Resources
- 📺 Watch the video: No Code Power BI — Claude and Power BI Modeling MCP Server
- 📖 Power BI Date Table: DAX or Power Query?
- 📖 All-in-One DAX Script: Create a Calendar/Date Dimension in Power BI
- 📖 Basics of Time Intelligence in DAX: YTD, QTD, MTD
- 📖 Power BI DAX: Previous MTD, QTD, and YTD Calculations
- 📖 Power BI File Formats Explained: PBIX vs PBIT vs PBIP vs PBIR
- 📖 Star Schema and Power BI Modeling Basics
- 🎓 RADACAD Training Courses — Power BI and Microsoft Fabric
Reza Rad is a Microsoft Regional Director, Microsoft Data Platform MVP, Author, Trainer, Speaker, and Consultant. He is co-founder of RADACAD and has more than 20 years of experience in data analytics, Power BI, and Microsoft BI. Connect with Reza on LinkedIn.
Reza is author of more than 14 books on Microsoft Business Intelligence, most of these books are published under Power BI category. Among these are books such as Power BI DAX Simplified, Pro Power BI Architecture, Power BI from Rookie to Rock Star, Power Query books series, Row-Level Security in Power BI and etc.
He is an International Speaker in Microsoft Ignite, Microsoft Business Applications Summit, Data Insight Summit, PASS Summit, SQL Saturday and SQL user groups. And He is a Microsoft Certified Trainer.
Reza’s passion is to help you find the best data solution, he is Data enthusiast.
His articles on different aspects of technologies, especially on MS BI, can be found on his blog: https://radacad.com/blog.