Expanding your config

Introduction

OK, we have spent a lot of time with tooling, with configuring app registrations and managed identity for the user context and CI/CD scenarios, plus working with repos and pipelines.

You are set to develop faster with a Git repo in the user context and then switch to the productionised version in the cloud Git provider and deployment.

This page covers some useful tooling and additional links to help you on your way as you expand your Terraform config in the fabric ecosystem.

Fabric Terraform provider documentation

Resource number one is the provider documentation itself. Terraform on Microsoft Fabric Documentation

Again, each resource and data source page will have information on whether the feature is preview or not, and whether it can be used by a service principal.

Let’s look at an example for Copy Jobs.

Many of the attributes are pretty simple and well described in the documentation and examples, but some of the more complex attributes - such as the definition here - are not well documented.

So how would you find out what the content of the definition JSON template should look like? Let’s see how MCP servers and the Fabric CLI can help us get there.

Model Context Protocol (MCP)

VS Code now supports MCP servers as first class resources and will either store the server config within an mcp.json in your roaming profile, or you can add that config as a .vscode/mcp.json file in your workspace.

In this section you will add the Microsoft Docs MCP server into your profile settings via the GUI, and then add the Terraform MCP server into your workspace. Once configured then we will run through an example set of prompts for GitHub Copilot.

Microsoft Docs MCP server

These steps are correct as of VS Code v1.102.1.

1. Extensions (CTRL+SHIFT+X) > MCP Servers > MCP Servers for agent mode

   

2. Click on Install Microsoft Docs

3. Click on Install

   

Remote http MCP servers are the simplest to set up.

Terraform MCP server

⚠️ Note that this MCP server is in beta at the time of writing.

Docker

The Terraform MCP server needs docker to be installed and running.

You can test the MCP server will work at the terminal by using:

docker run --interactive --rm hashicorp/terraform-mcp-server

Once docker downloads the image then you will see the following message.

HCP Terraform MCP Server running on stdio.

Use CTRL+D to exit and remove the container.

Configure

You could also add the Terraform MCP server config into the roaming mcp.json profile (%USERPROFILE%/AppData/Roaming/Code/User/mcp.json) alongside the definition for the Microsoft Learn Docs MCP server, but we will create a workspace specific file instead so that you can see how to control which servers are available for which projects.

1. Open the Explorer tab in Visual Studio Code (CTRL+SHIFT+E)

2. Add .vscode/mcp.json

   {
       "servers": {
           "terraform": {
               "type": "stdio",
               "command": "docker",
               "args": [
                   "run",
                   "--interactive",
                   "--rm",
                   "hashicorp/terraform-mcp-server"
               ]
           }
       }
   }
   

This approach is common as there is a limit of 128 tools in the GitHub Copilot agent mode and you may need to be selective. As an example, you may add an npm MCP servers for a Node.JS workspace, but you don’t need that generally. (The Microsoft Docs MCP server only uses one tool.)

Example Prompts

You are now ready to use the MCP server right within the IDE. Here are a few example prompts for you to test with.

1. Describe a resource

   Describe the fabric_copy_job resource in the microsoft/fabric provider
   

   You may have to approve the use of the MCP server first time round and then you should see a similar response to the screenshot below.

   

2. Describe a valid definition for the resource

   Use Microsoft Learn to describe a valid definition for this resource
   

   Add in the Microsoft Learn MCP server to get more definition.

   

   The MCP server has linked to an anchor within the copy job documentation for more detail.

3. Create an example template

   Generate a valid copyjob-content.json template including the token handlebars
   

   Agent mode can go further, generating files for you. Here it is creating an example template file for you to use as a starting point.

   

Using GitHub Copilot Agent mode in combination with the MCP servers is a powerful combination to get you working faster. We have a valid template, but it is just an example. What if you have a manually created resource and you need to query it?

Using the Fabric CLI

This is where the Fabric CLI is absolute gold dust. A common workflow is to manually create a customised resource in the Fabric portal, then use the Fabric CLI to inspect and understand its configuration. This approach helps you bridge the gap between manual changes and infrastructure as code.

1. Create the resource manually

   In the Fabric portal, create your custom resource, such as a Copy Job and configure it as needed.

2. Authenticate with the Fabric CLI

   Authenticate into Fabric.

   fab auth login
   

   Check your authentication status.

   fab auth status
   

3. List the workspaces

   fab ls
   

   Example output:

   My workspace.Personal
   aidwfactoryStreaming.Workspace
   aidwiotfactory.Workspace
   

4. List the resources in a workspace

   Example:

   fab ls aidwiotfactory.Workspace
   

   Example output:

   DataflowsStagingLakehouse.SemanticModel
   DataflowsStagingLakehouse.SQLEndpoint
   DataflowsStagingLakehouse.Lakehouse
   DataflowsStagingWarehouse.SemanticModel
   DataflowsStagingWarehouse.Warehouse
   IoT.Dashboard
   Test.Report
   Test OneLake.Report
   Test OneLake_Dataset_2080000_651f1831-8e0a-f011-bae2-000d3a3012e8.SemanticModel
   Test_Dataset_2080000_a508f593-770a-f011-bae2-000d3a3012e8.SemanticModel
   aidwBronze2SilverActivityData.DataPipeline
   aidwCopySilverActivity.CopyJob
   aidwFactoryBronze2SilverCompletions.DataPipeline
   aidwSilverToGold.Notebook
   aidwfactorybronze.SemanticModel
   aidwfactorybronze.SQLEndpoint
   aidwfactorybronze.Lakehouse
   aidwfactorygold.SemanticModel
   aidwfactorygold.SQLEndpoint
   aidwfactorygold.Lakehouse
   aidwfactorysilver.SemanticModel
   aidwfactorysilver.SQLEndpoint
   aidwfactorysilver.Lakehouse
   testSources.Eventstream
   

5. Inspect a resource

   Use the Fabric CLI to view a resource as JSON.

   Example:

   fab get aidwiotfactory.Workspace/testSources.Eventstream -q .
   

6. List queryable properties

   List the queryable properties in the resource:

   fab get aidwiotfactory.Workspace/testSources.Eventstream -f
   

   Example output:

   id
   type
   displayName
   description
   workspaceId
   definition.parts[0].path
   definition.parts[0].payload.sources
   definition.parts[0].payload.destinations
   definition.parts[0].payload.streams
   definition.parts[0].payload.operators
   definition.parts[0].payload.compatibilityLevel
   definition.parts[0].payloadType
   definition.parts[1].path
   definition.parts[1].payload.retentionTimeInDays
   definition.parts[1].payload.eventThroughputLevel
   definition.parts[1].payloadType
   definition.parts[2].path
   definition.parts[2].payload.$schema
   definition.parts[2].payload.metadata
   definition.parts[2].payload.config
   definition.parts[2].payloadType
   connections
   

   Note that the -f forces the output, otherwise you are prompted in v1.0.0 with:

   ? Item definition is retrieved without its sensitivity label. Are you sure? (Y/n)
   

7. Query a single property and create a JSON file

   There are a couple of small challenges with creating output for Terraform use:

   • the warning message above is also sent to stdout
   • the end of line is Windows format, i.e. CRLF

   This shell command takes the definition property and outputs to a file.

   fab get aidwiotfactory.Workspace/testSources.Eventstream -fq definition | grep -v "sensitivity label" | jq . > eventstream.json.tmpl
   

   You will then need to configure and test from there, but this approach should take you a long way.

Additional resources can be found here:

• https://github.com/RuiRomano/fabric-cli-powerbi-cicd-sample
• https://github.com/murggu/fab-demos

Git Integration with Microsoft Fabric

Terraform is not the only way to integrate Git with a Microsoft Fabric config.

Another option is to look at the Git Integration in Microsoft Fabric, enabling version control, collaboration, and CI/CD workflows for items within a Fabric workspace. This is handled directly within the Fabric GUI and is in preview at the time of writing.

Additional resource links

• https://microsoft.github.io/fabric-cli/
• https://github.com/RuiRomano/fabric-cli-powerbi-cicd-sample
• https://github.com/murggu/fab-demos

Next

These labs should have given you a good overview on how you may approach automating Microsoft Fabric deployments and configurations with the growing ecosystem of tooling including the Fabric CLI, Fabric Terraform provider, Git integration for Fabric and the Terraform MCP server, plus clear guidance on configuring access when in user context and using managed identity as your workload identity.