Chat with Fabric Data Agents in Visual Studio Code

This post contains the results of my initial test to chat with Fabric data agents in Visual Studio Code. Which is now possible due to the fact that the ability to consume Fabric data agent as a model context protocol server in Visual Studio Code is now available. Providing various prerequisites are met.

Introducing this capability is a significant development. Because you can add multiple MCP servers to the same instance of Visual Studio Code. Not just MCP servers that connect to Fabric data agents, but other types as well. Including the Microsoft Learn MCP server that I mentioned in my previous post.

I decided to base this example on the popular FUAM monitoring solution in Microsoft Fabric. Because it is a great real-world example. Due to the fact that it allows you to ask questions about the state of your Microsoft Fabric tenant with natural language. Which is a fantastic addition to your governance story.

For the benefit of this post, I created a basic Data Agent based on the FUAM_Core_SM semantic model. Which is the main semantic model that gets created as part of the FUAM monitoring solution.

As you can see above, it currently has two agent instructions. One about how to return workspace details and another about what to do when asked about Git integration. In reality, you require more instructions to get answers to more complex questions. However, I wanted to keep it simple for this post.

Anyway, after publishing the data agent I went to the settings and copied the MCP server URL.

I then went into Visual Studio Code to setup the MCP server.

Configuring the MCP server for Fabric Data Agents in Visual Studio Code

I then went into Visual Studio Code to setup the MCP server. As per the instructions I created an “mcp.json” file. From there, I went through the wizard and by clicking the “Add Server button”. Providing my MCP server with a sensible name.

Once done, I clicked start within the JSON file to start my MCP server and authenticated with my Microsoft Entra credentials.

Afterwards, I opened GitHub Copilot Agent mode in Visual Studio Code. I then asked how many workspaces are configured with Git integration. Which returned the correct number.

I then asked the chat if it could confirm that all the configured workspaces use Azure DevOps as a Git provider. Which returned the below response.

I then went on to ask a few other basic questions. Including what semantic models exist in a particular workspace.

Of course, this is just a small sample of the possibilities. As I mentioned earlier in the post, you can get answers to more complex questions by adding additional instructions to the Data Agent.

Metadata for the Fabric Data Agent MCP server in Git repository

Since I was curious, I configured the workspace the data agent was in with Microsoft Fabric Git integration. To To check if the Git repository stores any information about the MCP server. I can confirm that at this moment in time no MCP server related metadata is stored in the Git repository.

Final words

As you can see, you can configure the chat with Fabric Data Agents in Visual Studio Code with relative ease. Providing all the prerequisites are met.

In my honest opinion. Configuring the ability to consume Fabric data agent as a model context protocol server in Visual Studio Code is the easy part. The hard part is making sure you are working with the right data and instructions. In order to apply your prompt engineering skills correctly when asking questions in natural language.

Looking forward to seeing how others apply this functionality ongoing. Plus, which other applications this capability will appear in. Because I suspect a fair few will be keen for it to work in Claude Desktop.