Skip to content

DAXPerformanceTunerMCPServer

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

DAXPerformanceTunerMCPServer

README.md

DAX Performance Tuner

Transform slow DAX queries into lightning-fast optimized code using AI assistance and expert knowledge.

⚠️ Important Disclaimer

Always test optimized DAX queries thoroughly before deploying to production. While this tool performs semantic equivalence checking to verify that optimized queries return the same results as the original, this validation is not comprehensive. The semantic checking:

  • Compares results for the specific query context tested
  • May not cover all edge cases or data scenarios
  • Cannot guarantee identical behavior across all possible filter contexts, user interactions, or data states

🎥 Watch It In Action

See the DAX Performance Tuner in action:

DAX Performance Tuner Demo

Click the image above to watch the demo video

⚡ Quick Setup

Prerequisites

Before you begin, install:

  • ✅ .NET SDK 8.0+ - Download here
    • Required for building the MCP server from source

Installation Steps

  1. Download the Distribution

    Option A: Download Pre-Packaged Zip (Recommended)

    • Navigate to tools\DAXPerformanceTunerMCPServer\ in the repository
    • Download the DAXPerformanceTunerMCPServer_YYYYMMDD.zip file
    • Extract to your preferred location

    Option B: Clone Full Repository

    • Click the green "Code" button → "Download ZIP"
    • Extract and navigate to tools\DAXPerformanceTunerMCPServer\
    • Use the included zip or run the files directly from this folder

  2. Run Setup

    • Double-click setup.bat (or run setup.ps1 in PowerShell)

    The setup will:

    • ✅ Validate .NET SDK 8.0+ installation
    • ✅ Restore NuGet packages and build the MCP server from C# source
    • ✅ Generate MCP configuration in .vscode/mcp.json

    Note: The only file needed to run the MCP server is the dax-performance-tuner.exe produced by the setup scripts. For security reasons, pre-built executables are not stored in the fabric-toolbox repository — instead, setup.bat / setup.ps1 compile the exe from source so you can verify exactly what you're running.

  3. Start the MCP Server in VS Code

    • Open VS Code in the extracted DAXPerformanceTunerMCPServer\ folder
    • Open the .vscode\mcp.json file
    • Click the Start button over the server name

  4. Use with GitHub Copilot Chat

    • Open Copilot Chat
    • Ask: "Help me optimize this DAX query"
    • The server will automatically provide optimization tools

📝 Configuration for Other MCP Clients

For Claude Desktop:

  1. Run Setup First

    • Extract the zip file to your preferred location
    • Double-click setup.bat (or run setup.ps1 in PowerShell)
    • This builds the MCP server from source
    • Wait for setup to complete successfully

  2. Add Configuration

    • Open your claude_desktop_config.json file
    • Add the following configuration (replace paths with your actual installation location):
    {
  "mcpServers": {
    "dax-performance-tuner": {
      "command": "C:\\path\\to\\DAXPerformanceTunerMCPServer\\src\\DaxPerformanceTuner.Console\\bin\\Release\\net8.0-windows\\win-x64\\publish\\dax-performance-tuner.exe",
      "args": ["--start"]
    }
  }
}

    Important:

    • Use absolute paths for command
    • Use double backslashes (\\) in the JSON paths
    • Replace C:\\path\\to\\ with your actual installation directory

  3. Start the Server

    • Save the claude_desktop_config.json file
    • The server will start automatically when Claude Desktop launches

  4. Reset Claude Desktop

    • After updating the config and running the server, reset Claude Desktop

For Other MCP Clients:

  • Command: {install_path}\src\DaxPerformanceTuner.Console\bin\Release\net8.0-windows\win-x64\publish\dax-performance-tuner.exe (absolute path)
  • Args: ["--start"]
  • Restart the MCP client after configuration changes

🛠️ Tools You Get

Tool What It Does
connect_to_dataset Smart connection tool - Auto-discovers datasets, searches desktop instances, or connects directly based on what info you provide. Works with Power BI Service workspaces AND local Desktop instances. Call with no parameters to discover desktop instances (no auth). Add location="service" to discover workspaces instead.
prepare_query_for_optimization Complete baseline setup: inline measures, execute baseline, get metadata & research
execute_dax_query Test optimization attempts with automatic baseline comparison
get_session_status Track your optimization progress, view session history, and get intelligent next step recommendations

🚀 2-Stage Optimization Workflow

Stage 1 - Connection: Connect to your Power BI dataset
Stage 2 - Comprehensive Optimization: Complete baseline analysis → Iterative optimization → Performance validation

What Makes It Powerful:

  • Research-Driven: Automatically retrieves targeted optimization articles based on query patterns
  • Evidence-Based: Deep performance trace analysis with Formula Engine vs Storage Engine breakdown
  • Workflow-Guided: Systematic 2-stage process ensures comprehensive optimization
  • Iterative: Multiple optimization rounds achieve compound performance improvements
  • Semantic Validation: Guarantees optimized queries return identical results to baseline

📦 What's Included

  • C# MCP Server (src/) - Complete implementation with 4 specialized tools, built from source during setup
  • NuGet Dependencies - ADOMD.NET, MSAL, and other packages restored automatically during build
  • Automated Setup Scripts - setup.bat and setup.ps1 handle building and configuration

Attribution & Credits

This project builds upon and references valuable DAX optimization knowledge from the community:

We are grateful for their contributions to the DAX community. See ATTRIBUTION.md for detailed third-party content attribution.

Important: This project uses dual licensing. See license section below for details.

License

This project uses dual licensing:

MIT License

Applies to:

  • C# MCP server (src/) excluding DaxTraceRunner component
  • Configuration files
  • Setup scripts
  • Documentation (except third-party content)

Microsoft Reciprocal License (Ms-RL)

Applies to:

  • DaxTraceRunner component (src/DaxPerformanceTuner.Library/Infrastructure/DaxTraceRunner.cs)
  • This component contains code derived from DAX Studio
  • See src/LICENSE-MSRL.txt for full license text

What this means for users:

  • ✅ You can freely use, modify, and distribute this project
  • ✅ Most of the project is under permissive MIT license
  • ⚠️ If you modify the DaxTraceRunner component, you must share those changes under Ms-RL
  • ℹ️ Ms-RL is an OSI-approved open source license that only requires reciprocal sharing of modifications to Ms-RL files

See ATTRIBUTION.md for complete third-party content attribution and licensing details.