Fabric API Discovery

Fabric API Discovery

Fabric MCP Server

The Fabric Pro-Dev MCP Server provides local-first API discovery and OneLake management without connecting to live Fabric environments.

Setup

npm install -g fabric-pro-dev-mcp-server

Requires Node.js 18+.

MCP Configuration

Add to your .mcp.json or agent MCP config:

{
  "mcpServers": {
    "fabric-pro-dev": {
      "command": "npx",
      "args": ["-y", "fabric-pro-dev-mcp-server"]
    }
  }
}

Power BI MCP (Live Queries)

For live DAX queries against semantic models:

{
  "mcpServers": {
    "PowerBIQuery": {
      "type": "http",
      "url": "https://api.fabric.microsoft.com/v1/mcp/powerbi",
      "oauthClientId": "<MICROSOFT_PUBLIC_CLIENT_ID_FROM_CURRENT_DOCS>",
      "oauthPublicClient": true
    }
  }
}

Use the Microsoft-documented public client ID from the current Fabric MCP documentation rather than hardcoding a GUID in this repository.

API Discovery Tools

OneLake Tools

Available Workloads

Workloads with public API specifications include:

• Lakehouses
• Warehouses
• Notebooks
• Spark Job Definitions
• Data Pipelines
• Semantic Models (Power BI)
• KQL Databases
• Eventhouse
• Real-Time Intelligence (Eventstreams)
• ML Models & Experiments
• Reports
• Environments

Item Definitions

Fabric items can be created/updated via definition-based workflows using base64-encoded file parts:

POST /v1/workspaces/<wsId>/items/<itemId>/updateDefinition[?updateMetadata=true]
{
  "definition": {
    "format": "<format>",
    "parts": [
      { "path": "<filePath>", "payload": "<base64>", "payloadType": "InlineBase64" }
    ]
  }
}

Use publicapis_bestpractices_itemdefinition_get to discover the expected format and parts for each item type.

Gotcha: Omitting ?updateMetadata=true silently ignores the .platform part.

Developer vs Consumer Patterns

Developers (REST + Protocol)

• Use REST APIs to create/manage artifacts
• Use protocol-specific connections for data access:
  • Spark/PySpark for Lakehouse data
  • ODBC/JDBC for Warehouse queries
  • XMLA/DAX for Semantic Models
  • KQL for Real-Time Intelligence

Consumers (MCP)

• Use MCP servers for natural-language queries
• Supported for: Semantic Models, Warehouses, Lakehouse SQL Endpoints
• No ODBC/JDBC setup needed — MCP handles connections

CLI Tooling