Bulk edits across a semantic model: names, descriptions, translations.

Repetitive model upkeep is where it pays off: documenting every measure, applying one naming pattern across hundreds of objects or translating a whole model, edited directly in your Fabric workspace.

What it is

The Microsoft-hosted version of the Power BI Authoring MCP server, in preview. It lets an AI assistant create and change Power BI semantic models in Fabric workspaces, acting with your permissions. Its tools are grouped by object type, such as tables, columns, measures and relationships. A local version, generally available, also reaches Power BI Desktop and project files on your machine.

What you get

Tables, columns, measures, relationships, hierarchies, calculation groups, perspectives, partitions and security roles created, changed or deleted
Bulk changes across hundreds of objects: renames, refactoring, translations, security rules
A model checked against modeling best practices, with the fixes applied
DAX queries run to test measures and troubleshoot calculations while you build

Requirements

Your Fabric administrator must first turn on the tenant setting "Users can use the Power BI Model Context Protocol server endpoint".
You need Write permission on the semantic model to change it; with Build permission only, it can just run DAX queries.
You sign in with your Microsoft Entra account the first time the assistant uses it.
For Claude Code, you or a tenant admin first register your own Microsoft Entra app; it needs permission to change your semantic models, which some tenants require an admin to approve. The Install tab lists the steps.
Some edits cannot be undone; Microsoft advises backing up the model first.

Setup effort

Follow the published steps