Summary

The Power BI Modeling MCP Server bridges AI agents with Power BI semantic models, enabling natural language-driven development and management. It targets developers and AI application builders, offering automated modeling, bulk operations, and DAX validation to accelerate and enhance Power BI development workflows.

How It Works

This project implements the Modeling Capabilities Protocol (MCP) to expose Power BI semantic modeling functionalities to AI agents via a local server. It translates natural language commands into Power BI model operations (create, update, delete tables, measures, relationships, etc.), facilitating agentic workflows and complex model manipulations through a standardized interface.

Quick Start & Requirements

• Installation: Recommended via Visual Studio Code extension (requires VS Code, GitHub Copilot, and Copilot Chat). Manual installation involves downloading a .vsix file, unzipping, and configuring an MCP client.
• Prerequisites: VS Code, GitHub Copilot/Chat extensions. Connection to Power BI Desktop, Fabric workspace, or Power BI Project (PBIP) files.
• Links: An end-to-end demo video is available.

Highlighted Details

• Natural language interface for building and modifying semantic models (tables, columns, measures, relationships).
• Scalable bulk operations with transaction support for renaming, refactoring, and translations.
• Automated application of modeling best practices and agentic development workflows using TMDL and PBIP files.
• DAX query validation and execution for troubleshooting and data exploration.
• Extensive toolset covering database, model, table, column, measure, relationship, security, and calculation group operations.

Maintenance & Community

The project is in Public Preview, indicating active development. Feedback and issue reporting are encouraged via the public GitHub repository.

Licensing & Compatibility

• License Type: Not explicitly stated in the README.
• Compatibility: Can be configured for Analysis Services databases (--compatibility Full). Follows external tool modeling operation rules.

Limitations & Caveats

Currently in Public Preview, subject to significant changes before General Availability. Potential for LLM-generated inaccuracies or unexpected results; sensitive data exposure in logs is a risk. Model backups are strongly advised. The server is limited to modeling operations; it does not modify report pages or diagram layouts. Performance and relevance are highly dependent on the chosen AI model. Fabric workspace connection may have tenant-specific rollout issues. Requires careful security review of AI clients and RBAC permissions due to potential for destructive actions.