PlugWiki
Power BI logo

Power BI MCP servers: which one to use

Use Microsoft's Fabric IQ to ask questions about your data, and its Authoring server to edit models. Community servers add read-only and Desktop options.

Servers
15
active
Official
1
by Power BI
Local
9
runs on your machine
OAuth
1
sign in, no key

Microsoft has official servers for Power BI, and they do different jobs. Fabric IQ answers questions about your data, and the Authoring server builds and changes semantic models. Microsoft says not to use the Authoring server for everyday questions. Authoring can delete model objects, so use its read-only flag or back up the model first.

Good for

  • Analysts who want to ask questions of a Power BI semantic model in plain English
  • Model builders who want an assistant to add measures, rename objects and apply best practices
  • Teams who want an assistant to read data without changing the model
  • Developers working on Power BI Project files in source control

Which Power BI MCP server to use

Use Fabric IQ for questions and Power BI Authoring for building models, since both come from Microsoft. Pick Power BI Analyst for a local, read-only server, or the community Power BI MCP Server for Desktop work with a read-only switch.

ServerOfficialRunsSign-inRead-only optionGitHub stars
For Power BI
Power BI AuthoringThis pageOfficialHostedLocalAccountFlag1.3K
Power BI MCP ServerCommunityLocalService principalEnv setting125
SemanticOpsCommunityLocal––255
Power BI AnalystCommunityLocalAccountAlways read-only14
Other tools for the same job
Fabric IQOfficialHosted–––
Power BI remote (earlier)OfficialHostedAccount––

All Power BI MCP servers

15 active · 1 official · official first, then by GitHub stars

Questions

Is there an official Power BI MCP server?

Yes. Microsoft documents two: the Power BI Authoring server, which builds and edits semantic models, and Fabric IQ, which answers questions about your data. An earlier Power BI query endpoint is still documented for existing setups.

Which Power BI MCP server should I use to ask questions about my data?

Use Fabric IQ. Microsoft says not to use the Authoring server for that, because it is built for modeling. For a local, read-only option, the community Power BI Analyst server runs DAX queries against your semantic models.

Can a Power BI MCP server edit my model?

Yes. The Authoring server can create, update and delete tables, columns, measures, relationships and security roles. Its README warns that the AI can make unintended changes, so back up your model first.

Which Power BI MCP servers are read-only?

Power BI Analyst only reads. Microsoft's local Authoring server has a --readonly flag, and the community Power BI MCP Server refuses writes when you set POWERBI_MCP_READONLY=true.

Is there a hosted Power BI MCP server, and what is its URL?

Yes. The Authoring server is hosted by Microsoft at https://api.fabric.microsoft.com/v1/mcp/powerbi/authoring, in preview. Fabric IQ is at https://fabriciq.svc.cloud.microsoft/v1/mcp/FabricIQ. The local Authoring server is generally available and adds Power BI Desktop and project files.

Does a Power BI MCP server work with Claude?

Yes. Microsoft's local Authoring server can be added to any MCP client through its npx setup. The Power BI Analyst README covers Claude Desktop, and the SemanticOps README lists Claude Desktop and Claude Code.

Does my Power BI admin need to approve anything?

Sometimes. The earlier Power BI endpoint needs an admin to turn on a tenant setting for the preview endpoint. Power BI Analyst needs the Dataset Execute Queries REST API setting on. You also need Build permission on the semantic model.

Do these servers work on a Mac?

SemanticOps lists Windows 10/11 and macOS. The community Power BI MCP Server connects live to Power BI Desktop only on Windows, and its offline file tools run on any system.

What licenses do they use?

Microsoft's npm Authoring server requires you to accept its EULA before the tools run. The community Power BI MCP Server and Power BI Analyst are MIT licensed. SemanticOps has a proprietary license that allows personal and commercial use.

Server facts from the official MCP registry, GitHub, npm, PyPI and the Docker catalog. Archived servers are not listed.