Power BI MCP Server

An MCP (Model Context Protocol) server that connects to Power BI semantic models, enabling natural language queries via DAX through AI assistants like Claude.

Features

• List workspaces — Browse all Power BI workspaces you have access to
• List datasets — See semantic models in any workspace
• Discover model schema — Inspect tables, columns, measures, and data types via DMV queries
• Execute DAX queries — Run DAX EVALUATE statements and get structured results
• Refresh datasets — Trigger and monitor dataset refreshes
• Get dataset details — View metadata for any semantic model

Authentication

Authenticates with Microsoft Entra ID, following the same patterns as the Azure MCP and Azure DevOps MCP servers. Three modes are supported:

| Mode | Flag | Description | |------|------|-------------| | Interactive (default) | --authentication interactive | Opens a browser for OAuth login | | Azure CLI | --authentication azcli | Uses your az login session | | Environment | --authentication env | Uses DefaultAzureCredential (service principal env vars, managed identity, etc.) |

You can optionally pass --tenant <tenant-id> for single-tenant or B2B scenarios.

Setup

Prerequisites

• Node.js 18+
• An Azure/Microsoft Entra ID account with access to Power BI

Build

npm install
npm run build

Configure in Claude Code

Add to your Claude Code MCP settings (~/.claude/settings.json or project .claude/settings.json):

{
  "mcpServers": {
    "powerbi": {
      "command": "node",
      "args": ["/path/to/PowerBIMCP/dist/index.js", "--authentication", "interactive"],
      "env": {}
    }
  }
}

For Azure CLI authentication (if you're already logged in via az login):

{
  "mcpServers": {
    "powerbi": {
      "command": "node",
      "args": ["/path/to/PowerBIMCP/dist/index.js", "--authentication", "azcli"],
      "env": {}
    }
  }
}

With a specific tenant:

{
  "mcpServers": {
    "powerbi": {
      "command": "node",
      "args": [
        "/path/to/PowerBIMCP/dist/index.js",
        "--authentication", "interactive",
        "--tenant", "your-tenant-id"
      ],
      "env": {}
    }
  }
}

Available Tools

list_workspaces

List all Power BI workspaces the authenticated user has access to.

list_datasets

List datasets (semantic models) in a workspace.

• workspaceId (optional) — omit for "My Workspace"

get_dataset

Get detailed information about a specific dataset.

• datasetId (required)
• workspaceId (optional)

get_tables

List tables in a dataset via the REST API.

• datasetId (required)
• workspaceId (optional)

discover_model

Discover the full schema of a semantic model (tables, columns, measures) using DMV queries. This is the best tool to call first before writing DAX queries.

• datasetId (required)
• workspaceId (optional)

execute_dax_query

Execute a DAX query against a semantic model.

• datasetId (required)
• query (required) — a valid DAX EVALUATE statement
• workspaceId (optional)

refresh_dataset

Trigger a dataset refresh.

• datasetId (required)
• workspaceId (optional)

get_refresh_history

Get refresh history for a dataset.

• datasetId (required)
• workspaceId (optional)
• top (optional) — number of entries, defaults to 10

Example Workflow

1. List workspaces to find your workspace ID
2. List datasets in that workspace to find your semantic model
3. Discover model to understand the schema (tables, columns, measures)
4. Execute DAX queries to answer questions about your data

"Show me total sales by region for 2024"
→ discover_model → execute_dax_query with:
  EVALUATE SUMMARIZECOLUMNS('Geography'[Region], 'Date'[Year], "Total Sales", SUM('Sales'[Amount]))

Debugging

Set LOG_LEVEL=debug for verbose logging:

{
  "mcpServers": {
    "powerbi": {
      "command": "node",
      "args": ["/path/to/PowerBIMCP/dist/index.js"],
      "env": {
        "LOG_LEVEL": "debug"
      }
    }
  }
}

Use the MCP Inspector for interactive testing:

npm run inspect