Teaching Claude and Copilot to speak Power BI

What it does

This repo is a plugin marketplace that teaches Claude Code and GitHub Copilot how to work with Power BI and Microsoft Fabric. It packages domain knowledge—DAX, TMDL, PBIP formats, Tabular Editor workflows—into installable skills, autonomous agents, and validation hooks. The goal is to let an AI agent generate, review, and catch errors in semantic models, reports, and Fabric resources instead of doing it all by hand.

The interesting bit

The project treats enterprise BI metadata as code that coding agents can manipulate, and it uses deterministic hooks to validate file changes automatically rather than trusting the LLM to self-correct. It also straddles both Anthropic’s Claude Code and GitHub’s Copilot CLI using a shared manifest format, though the README is frank about where the two tools diverge.

Key highlights

• Covers the full Power BI/Fabric stack: semantic models, TMDL, PBIR, DAX, Tabular Editor, and Fabric CLI
• Includes autonomous subagents for tasks like BPA rule debugging and PBIP project validation
• Hooks run deterministic checks after tool use, such as validating measure metadata, referential integrity, and TMDL syntax
• Dual-targeted: works as a Claude Code plugin marketplace and Copilot CLI plugin source
• Weekly release cadence with active development

Caveats

• Under active development with regular renaming or restructuring; expect churn
• Windows users may hit MAX_PATH limits with TMDL files unless long-path support is enabled
• Some hook features require Copilot CLI ≥ 1.0.26; older builds resolve paths incorrectly

Verdict

Worth exploring if you manage Power BI or Fabric assets and want an AI agent to handle the boilerplate of DAX, TMDL, and report validation. Skip it if you aren’t using Claude Code or Copilot CLI, or if you need stable, frozen APIs.