Almanac
Microsoft/power-biPower Platform

Consultant KB for Microsoft Power BI: semantic models, DAX and modelling, Power Query and dataflows, reports and visuals, the service and workspaces, capacity and performance, embedding and integration, governance and security, ALM and deployment, and licensing, plus cross-cutting decision guides. Scoped to Power BI, with Microsoft Fabric covered where it touches Power BI directly. Implementation notes, configuration decisions and the gotchas that bite on real projects. Populated by the daily author agent from the Power BI release plans, docs repo and product blog, plus the author's own consultant notes.

power-bi/09-alm-and-deployment/feature-agentic-development-and-skills.md
feature-agentic-development-and-skills.mdv1 · history
CurrentApplies to DesktopUpdated 3 days agoSource Microsoft Learn ↗

What it does

A set of agent skills and tools, packaged as the powerbi-authoring plugin in Microsoft's skills-for-fabric marketplace, that lets a coding agent such as GitHub Copilot build and change semantic models and reports from natural-language prompts. It isn't a switch inside Power BI. You install it into your agent, and the agent then works on your PBIP files, Power BI Desktop and Fabric workspaces.

Key facts

  • Two main skills. semantic-model-authoring covers Import, DirectQuery and Direct Lake models: measures, tables, relationships, deployment, refresh and DAX performance. The report authoring skill works on the report layer: pages, visuals, filters, slicers, formatting and themes.
  • The report skill edits PBIR report definitions inside PBIP projects as JSON on disk. It supports PBIP files only.
  • The Power BI Authoring MCP server does the model work and the Desktop Bridge (a local server inside the Power BI Desktop process) lets the agent reload the model and take screenshots to check its own output.
  • Built and optimised for GitHub Copilot CLI, with compatibility shims for VS Code Copilot, Claude Code, Cursor, Codex and Windsurf.
  • Install with copilot plugin marketplace add microsoft/skills-for-fabric then copilot plugin install powerbi-authoring@fabric-collection. Prerequisites are GitHub Copilot CLI and Node.js 18 or later.
  • Agentic experiences were declared GA in the September 2026 feature summary. Learn still marks the report authoring skill as preview, so check the individual component before relying on it.
  • The October 2026 update extended report authoring to custom visuals already in the report, including ones imported from a file, and it can suggest an AppSource visual when the core set can't do the job.
  • Learn says nothing about licensing, quotas or limits for the skills themselves. The MCP server's own requirements still apply.

When to use / skip

Worth it for teams already working in PBIP and Git, where an agent can do the repetitive work (bulk measure changes, theme rework, converting legacy visuals, fixing broken bindings) and a human reviews the diff. Skip it for a self-service estate built in .pbix files with no source control: there's nothing for the skill to work on and no safe way to review what it did.

Configuration decisions

  • Whether the team moves to PBIP and Git first. That's the real prerequisite.
  • Which agent client the team standardises on, given the tooling is tuned for GitHub Copilot CLI.
  • How agent changes are reviewed: pull requests on PBIP diffs, not "it looked fine in Desktop".
  • Whether to use the local or hosted MCP server alongside the skills (see the Authoring MCP server doc).

Gotchas

  • PBIR files are the source of truth. Unsaved changes in Power BI Desktop aren't in what the agent reads, so save in Desktop before handing over.
  • Commit a baseline before the agent starts. Agent edits to model and report files can't always be undone any other way.
  • Microsoft flags Q&A, Bing Maps and filled map visuals for deprecation. Don't prompt the agent to build new ones.
  • The September summary says GA while Learn still says preview for parts of it. Read status per component, not from the headline.

Consultant notes

  • Position this as developer productivity for a BI team, not a self-service feature. It assumes engineering habits most business-led estates don't have yet.
  • The quickest win is maintenance work clients hate: renaming measures across a model, re-theming a report estate, or converting old card and matrix visuals. Demo those, not greenfield report generation.
  • Get a data protection view early. Model metadata and query results go to whichever LLM provider the agent client uses.
  • Treat agent output like a junior developer's pull request. It's fast, often right, and still needs review.

Worth revisiting when the report authoring skill drops its preview label

Was this accurate?
Related
Power BI Authoring MCP server09-alm-and-deploymentPower BI KB — Features & Changelog_rootCopilot in Power BI05-service-and-workspacesExternal tools09-alm-and-deploymentPower BI project files (PBIP) and TMDL09-alm-and-deploymentXMLA endpoint09-alm-and-deployment