Power BI Desktop MCP 1.3 - Public Release

After a closed beta and a round of bug-fixes, v1.3.0 is now available for everyone.

This project is now SemanticOps MCP (formerly MCP Engine). The original release artifacts remain available through the same GitHub releases page.

It’s an MCP server for Power BI Desktop that lets AI assistants (e.g., Claude) see and operate your local Power BI model: list tables/columns/measures, run DAX, create/update measures, manage relationships, inspect VertiPaq stats, and analyze DAX performance - all from a chat.

Why it matters

• Turn natural language into concrete Power BI model actions.

• Automate routine modeling tasks and diagnostics.

• Keep data local: the server runs on your machine and does not phone home.

Highlights

• 25 tools across model, table, column, measure, relationship, search, and advanced analysis.

• Local-only bridge between Power BI Desktop and your AI client.

• Three install options: Claude Desktop .mcpb (recommended), or .exe for Claude Desktop / Claude Code.

Install (2 minutes)

1. Download the release: .mcpb or PbiMcp.exe

2. For Claude Desktop, double-click the .mcpb and restart Claude.

   Or configure the .exe in claude_desktop_config.json.

3. Open a PBIX in Power BI Desktop. Ask Claude: “List my available Power BI models,” then “Connect to model 1.”

Try these prompts

• “Show all tables and measures in this model.”

• “Run DAX: EVALUATE TOPN(10, Sales, [SalesAmount], DESC).”

• “Create a Total Revenue measure with SUM(Sales[Amount]) and percentage format.”

• “List relationships and flag any bi-directional or many-to-many.”

• “Show VertiPaq stats; recommend columns to reduce cardinality.”

Limits (read before using)

• Windows only. Requires Power BI Desktop with an open PBIX.

• Local Analysis Services session; default DAX result cap 1000 rows.

• Not affiliated with Microsoft.

Privacy & licensing

• The MCP server does not collect or transmit telemetry or model data.

• Your AI client may retain chat history per its own policies.

• Proprietary license; redistribution/modification is not permitted.

Roadmap

I’m already testing V2 with broader capabilities and quality-of-life improvements. If you have specific workflows you want to automate, tell me.

Get it

➡️ Download: https://github.com/maxanatsko/mcp-engine-public/releases