Integration · Microsoft Power BI

Connect Microsoft Power BI to meinGPT

Find workspaces and semantic models, inspect model schema, and query measures through DAX in the signed-in user's context.

The Power BI integration consists of two coordinated connectors. Power BI Discovery finds workspaces, semantic models, reports, pages, and dashboards with their IDs. Microsoft's official Power BI MCP server reads a selected model's schema and executes DAX queries. Both are read-only and use the signed-in user's permissions.

Discovery uses Microsoft Graph to resolve the right Power BI resources. For data queries, meinGPT connects to Microsoft's official remote MCP endpoint through OAuth. The assistant must first read tables, columns, measures, and relationships, then write DAX only against the actual schema. Model queries require Build permission, and Row-Level Security is never bypassed.

Compare connection pathsThe same system can be connected differently depending on the task. Freshness, permissions, and scope determine the right path.
Microsoft Power BIDiscoverymeinGPT

Find content and resource IDs

Lists workspaces, semantic models, reports, report pages, and dashboards so the assistant can select the correct resource.

Best for
Orientation, resource selection, and preparation for a model query
Freshness
Live for each request
Permissions
Permissions of the signed-in Power BI user
Setup in the docs
Microsoft Power BIMicrosoft MCPmeinGPT

Query semantic models

Reads tables, measures, and relationships and executes controlled DAX queries through Microsoft's official endpoint.

Best for
Measures, ad-hoc analysis, and reviewable DAX queries
Freshness
Live against the selected semantic model
Permissions
Workspace access, Build permission, and Row-Level Security
Setup in the docs

Source, permissions, and data residency are documented separately for each connector.

Typical workflows

  • Find the correct semantic model
  • Ask for measures in natural language
  • Explain model and report structure

Compare connection paths

The same system can be connected differently depending on the task. Freshness, permissions, and scope determine the right path.

Discovery

Find content and resource IDs

Lists workspaces, semantic models, reports, report pages, and dashboards so the assistant can select the correct resource.

Best for
Orientation, resource selection, and preparation for a model query
Access
Read-only through Microsoft Graph
Freshness
Live for each request
Permissions
Permissions of the signed-in Power BI user
Scope
Workspaces and content visible to the connected account
Limitation
Discovery returns metadata and IDs, not measures from the semantic model.

Setup

  1. 1.Enable Power BI Discovery on the assistant
  2. 2.Connect the Microsoft account and select the resource
Setup in the docs

Microsoft MCP

Query semantic models

Reads tables, measures, and relationships and executes controlled DAX queries through Microsoft's official endpoint.

Best for
Measures, ad-hoc analysis, and reviewable DAX queries
Access
Read-only through the official Power BI MCP server
Freshness
Live against the selected semantic model
Permissions
Workspace access, Build permission, and Row-Level Security
Scope
One explicitly selected semantic model
Limitation
The Microsoft MCP endpoint is a preview and cannot change data, models, or reports.

Setup

  1. 1.Enable the MCP tenant setting through a Power BI administrator
  2. 2.Verify Build permission and connect the Microsoft account
Setup in the docs

What the integration makes possible

01

Find the correct semantic model

Use Discovery to resolve workspaces and models instead of copying or guessing resource IDs.

02

Ask for measures in natural language

Read the real schema, construct DAX from existing fields and measures, and answer from returned rows.

03

Explain model and report structure

Describe tables, measures, relationships, and report metadata without changing the model.

Clarify before starting

  1. 01The relevant tenant, workspaces, and semantic models are known.
  2. 02The Microsoft MCP preview is enabled in the tenant.
  3. 03Build permission and Row-Level Security have been reviewed.

Review the prompt and result openly

Prompt

Find the Sales semantic model in the Controlling workspace. Read its schema, then show revenue and contribution margin by region for the last completed quarter. Use existing measures and include the DAX query.

Result

The selected model and existing measures are named above a regional result table. The read-only DAX query is shown, and the connected user's Row-Level Security remains active.

Does this connection fit your use case?

We clarify the source, permissions, and connection required for the concrete use case. If the existing integration fits, you can move straight to setup afterwards.

Discuss the connectionStart for free

Security and operations

Both connectors run in the signed-in Microsoft user's context. Workspace rights, Build permission, and Row-Level Security remain unchanged, and the registered Microsoft MCP endpoint is read-only.

Known limitations

  1. 01The entire integration is read-only.
  2. 02Model queries require the tenant preview and Build permission.
  3. 03The assistant must inspect schema instead of guessing table or measure names.

Frequently asked questions

Yes. It reads the actual model schema, constructs DAX, and executes the query read-only against the selected semantic model.

Continue reading