Get started with Fabric Core MCP Server

Connect your AI agent to Microsoft Fabric using the cloud-hosted Core MCP Server and start managing workspaces with natural language. This quickstart guides you through setup and your first interaction.

Time to complete: 5-10 minutes

Note

Fabric Core MCP Server is currently in preview. Features and configuration may change before general availability.

Tip

Want to understand the architecture and security model first? See the Fabric Core MCP Server overview. Or compare with Fabric Pro-Dev MCP Server for local development.

Prerequisites

Before you begin, ensure you have:

Microsoft Fabric access — An active Fabric tenant with at least one workspace you can access
MCP-compatible agent — VS Code with GitHub Copilot (recommended) or Claude Desktop
Microsoft Entra ID account — Your Microsoft work or school account with Fabric permissions

Add Fabric Core MCP Server

Select your AI agent to see setup instructions:

Configure VS Code

In VS Code, press Ctrl+Shift+P (or Cmd+Shift+P on macOS) to open the command palette.
Type MCP: Add Server and select HTTP from the list.

Enter the Fabric Core MCP Server URL:

https://api.fabric.microsoft.com/v1/mcp/core

When prompted for a server name, enter fabric.
VS Code opens your browser for authentication. Sign in with your Microsoft Entra ID account.

The browser window closes automatically after successful authentication.

Alternative: Manual configuration

You can also configure Fabric Core MCP Server by editing your VS Code settings file directly.

Add the following to .vscode/mcp.json:

{
  "servers": {
    "fabric": {
      "type": "http",
      "url": "https://api.fabric.microsoft.com/v1/mcp/core"
    }
  }
}

Configure Claude Desktop

Locate your Claude Desktop configuration file:
- macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows: %APPDATA%\Claude\claude_desktop_config.json
Open the configuration file in a text editor.

Add the Fabric Core MCP Server configuration:

{
  "mcpServers": {
    "fabric": {
      "type": "http",
      "url": "https://api.fabric.microsoft.com/v1/mcp/core",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Replace YOUR_TOKEN with an access token.

Get an access token

To obtain an access token for Fabric:

Open a terminal or command prompt.

Run the following Azure CLI command:

az account get-access-token --resource https://api.fabric.microsoft.com

Copy the accessToken value from the response.
Paste it into your Claude Desktop configuration file replacing YOUR_TOKEN.

Important

Access tokens expire after a period of time (typically one hour). If authentication fails, generate a new token and update your configuration.

Generic MCP client configuration

For custom MCP clients or other AI agents that support MCP:

Connection details:

Property	Value
Endpoint URL	`https://api.fabric.microsoft.com/v1/mcp/core`
Auth type	OAuth 2.0 Bearer Token
Auth scope	`https://api.fabric.microsoft.com/.default`
Identity provider	Microsoft Entra ID

Implementation:

Configure your MCP client to connect to the Fabric Core MCP Server URL.
Implement OAuth 2.0 authentication flow using Microsoft Entra ID.
Include the access token in the Authorization header:
```
Authorization: Bearer {access_token}
```
Set the token scope to https://api.fabric.microsoft.com/.default.

Refer to your AI agent's MCP configuration documentation for specific setup steps.

Verify the connection

After configuring your AI agent, verify that Fabric Core MCP Server is connected:

In VS Code, open GitHub Copilot Chat by pressing Ctrl+Alt+I (or Cmd+Alt+I on macOS).
Enter the following prompt:
```
List all my Fabric workspaces
```
Copilot uses Fabric Core MCP Server to retrieve your workspaces.

You should see a list of workspace names that you have access to.

Open a conversation in Claude Desktop.
Enter the following prompt:
```
List all my Fabric workspaces
```
Claude uses Fabric Core MCP Server to retrieve your workspaces.

You should see a list of workspace names that you have access to.

Tip

If the connection fails, disconnect and reconnect the MCP server in your AI agent settings. Ensure you have at least Viewer access to one Fabric workspace in your tenant.

Try example prompts

Now that you're connected, try these common operations:

Explore resources:

Show me all items in the Sales Analytics workspace

What lakehouses exist in my development environment?

Create resources:

Create a workspace called Sales Analytics Dev

Create a lakehouse named CustomerData in the Dev workspace

Manage permissions:

List all role assignments for the Finance workspace

Add john@contoso.com as a Contributor to my Dev workspace

Optional: Add Microsoft Graph MCP integration

To enable user and group resolution by email address, you can also connect the Microsoft Graph MCP Server. This enhancement allows you to grant access using email addresses instead of looking up user IDs manually.

Why add Graph MCP

Without Graph MCP:

Add user {uuid-guid} as Contributor to Sales Analytics

With Graph MCP:

Add john@contoso.com as Contributor to Sales Analytics

The AI agent automatically resolves email addresses to user principal IDs.

Example commands with Graph MCP

"Add the Marketing team to Sales Analytics workspace as Viewers"
"Grant john@contoso.com Contributor access to my Dev workspace"
"Who are the members of the Finance Admins group?"

Setup Graph MCP

Follow the setup instructions at Get started with Microsoft MCP Server for Enterprise.

Clean up

To remove the Fabric Core MCP Server connection:

Press Ctrl+Shift+P (or Cmd+Shift+P on macOS).
Type MCP: Remove Server and select it.
Select fabric from the list.

Next steps

Last updated on 2026-04-13