PBIXRay MCP Server V2 - Enhanced Edition

Version: 2.0 Enhanced
Status: Production Ready Last Updated: 2025-10-04

What is This?

The PBIXRay MCP Server V2 is a Model Context Protocol (MCP) server that enables Claude AI to analyze Power BI Desktop models with advanced performance analysis capabilities including Storage Engine (SE) and Formula Engine (FE) breakdown.

Key Features

Auto-detect Power BI Desktop instances
Full Model Exploration - Tables, columns, measures, relationships
DAX Analysis - Measures, calculated columns, with expressions
Power Query (M) - View data source configurations
Performance Analysis - Advanced SE/FE timing with SessionTrace
VertiPaq Stats - Storage and compression metrics
Search Capabilities - Find objects and text across the model
Natural Language - Ask Claude questions about your model

System Requirements

OS: Windows 10/11 (64-bit)
.NET Framework: 4.7.2+ (usually pre-installed)
Power BI Desktop: Latest version recommended
Claude Desktop: Latest version
Disk Space: ~200 MB

Quick Start

1. Extract the Package

Extract the entire folder to your preferred location, for example:

C:\Tools\pbixray-mcp-server

2. Verify Installation (Optional)

cd C:\Tools\pbixray-mcp-server .\scripts\test_connection.ps1

3. Configure Claude Desktop

.\scripts\install_to_claude.ps1

This will automatically configure Claude Desktop for you.

4. Restart Claude Desktop

Fully close Claude Desktop (check Task Manager)
Reopen Claude Desktop

5. Test It!

Open Power BI Desktop with a .pbix file
In Claude, say: "Detect my Power BI Desktop instances"
Say: "Connect to instance 0"
Say: "What tables are in this model?"

You are now analyzing Power BI with Claude AI!

Documentation

Quick Reference - docs/QUICK_REFERENCE.md - Command cheat sheet
FAQ - docs/FAQ.md - Frequently asked questions
Troubleshooting - docs/TROUBLESHOOTING.md - Common issues
Deployment Guide - docs/DEPLOYMENT_GUIDE.md - Team deployment

Available Tools

Key tools available:

detect_powerbi_desktop - Find running Power BI instances
connect_to_powerbi - Connect to a specific instance
list_tables - List all tables in the model
list_measures - List DAX measures
describe_table - Get detailed table information
run_dax_query - Execute DAX queries
analyze_query_performance - SE/FE performance analysis
get_vertipaq_stats - Storage and compression metrics
search_objects - Find tables, columns, measures
export_model_schema - Export complete model structure

Folder Structure

pbixray-mcp-server/ ├── venv/ # Python environment (portable) ├── src/ # Server source code ├── lib/dotnet/ # Analysis Services DLLs ├── docs/ # Documentation ├── scripts/ # Helper scripts ├── config/ # Configuration templates ├── logs/ # Auto-generated logs └── requirements.txt # Python dependencies

Example Usage

You: "Analyze the performance of this DAX query" Claude: [Runs analyze_query_performance with 3 runs] Results: - Total: 245ms - Storage Engine: 189ms (77%) - Formula Engine: 56ms (23%) - SE Queries: 12 The query is SE-heavy, indicating most time is spent retrieving data. Consider adding filters to reduce the amount of data scanned.

Maintenance

Update Python Packages

.\venv\Scripts\pip.exe install --upgrade -r requirements.txt

Verify Health

.\verify_installation.ps1 -Verbose

For Team Distribution

To package for colleagues:

.\scripts\package_for_distribution.ps1

This creates a ready-to-share ZIP file on your Desktop.

Troubleshooting

No instances detected?

Ensure Power BI Desktop is running
Open a .pbix file (not just Power BI)
Wait 10-15 seconds after opening

Connection fails?

Verify instance index (usually 0)
Restart Power BI Desktop
Check docs/TROUBLESHOOTING.md

Claude does not see the server?

Run: .\scripts\install_to_claude.ps1
Restart Claude Desktop completely
Check JSON syntax in config

Credits

Built with:

MCP SDK - Model Context Protocol
Python.NET - CLR integration
Analysis Services Client Libraries - Microsoft
Power BI Desktop - Microsoft

Version History

V2.0 Enhanced (Oct 2025) - SessionTrace integration, improved SE/FE analysis
V2.0 (Oct 2025) - WMI-based detection, stability improvements
V1.0 (Sep 2025) - Initial release

Privacy and Security

All processing is local - No data sent to external servers
Localhost only - Binds to 127.0.0.1
No network exposure - Completely offline capable
Claude conversations - May be stored by Anthropic (review their policy)

Happy Analyzing!

For questions or issues, see the FAQ or Troubleshooting Guide in the docs folder.

This server cannot be installed

-

security - not tested

F

license - not found

-

quality - not tested

How are these scores calculated?

local-only server

The server can only run on the client's local machine because it depends on local resources.

Enables Claude AI to analyze Power BI Desktop models with comprehensive capabilities including DAX analysis, performance monitoring with Storage Engine/Formula Engine breakdown, VertiPaq statistics, and natural language querying of model structure and data.

PBIXRay MCP Server V2