# Microsoft Power BI MCP server

Source: https://agentic-cmms.klementine.xyz/integrations/microsoft-power-bi

> Let Claude, ChatGPT and other AI agents query Power BI semantic models for MTBF, MTTR, backlog and planned versus reactive work in plain language.

Microsoft Power BI is where many maintenance and reliability teams keep their KPIs. MTBF by asset class, MTTR by shift, backlog in weeks, PM compliance, planned versus reactive hours: someone built the semantic model, wrote the DAX measures and laid out the dashboards. Then everyone looks at the same five visuals every Monday.

The trouble starts with the sixth question. "Why did MTTR on Line 4 jump in September?" is not on any page, so it goes to the one person who knows the model. Connected through Agentic CMMS, an AI agent can ask the semantic model directly, using the measures you already trust, and explain what it finds.

## What your AI agent can do with Power BI

- **Read a KPI.** "What is MTBF for the filler machines this quarter compared with last?" The agent queries the measure in your semantic model and gives both values.
- **Explain a change.** "Why did MTTR on Line 4 rise in September?" It breaks the measure down by asset, failure code and shift to find the driver.
- **Check the backlog.** "How many weeks of maintenance backlog do we have at each plant?" You get one figure per plant, highest first.
- **Track planned versus reactive.** "What share of maintenance hours was planned work last month, by site?"
- **Find the right report.** "Which report shows PM compliance by technician?" The agent searches the workspaces you can see and gives you the report name.
- **Explain a measure.** "How is wrench time calculated in the reliability model?" It reads the DAX behind the measure and explains it in plain English.
- **Check data freshness.** "When did the maintenance dataset last refresh, and did it fail?" With write access, the agent can start a new refresh.

## Power BI with your other systems

Power BI shows the trend. Your source systems hold the detail behind it:

- When MTBF drops for a pump class, ask the agent to list the matching failure work orders in [Fiix](/integrations/fiix) or [SAP Plant Maintenance](/integrations/sap-plant-maintenance).
- Combine the backlog figure from Power BI with labour availability, then post a short weekly summary to [Microsoft Teams](/integrations/microsoft-teams).
- Go past the model and query the raw maintenance history in [Snowflake](/integrations/snowflake) when a question needs data the model does not hold.
- Set maintenance cost per asset from [Xero](/integrations/xero) next to downtime from Power BI to decide what to replace.

## One definition of every KPI

Maintenance KPIs are easy to calculate in slightly different ways, and that causes arguments. Because the agent asks your semantic model rather than calculating its own version, every answer uses the definitions your reliability engineers agreed on. Ask it how a figure was calculated and it shows you.

Each person approves their own agent with their own login, and the agent only sees the workspaces and models that connection allows. Owners and admins choose which roles may use each tool, and every query is written to the activity log.

## Questions and answers

### Can Claude answer questions from our Power BI semantic models?

Yes. The agent queries the semantic models you give it access to and uses the measures already defined there, so MTTR in a chat matches MTTR in your report.

### Will ChatGPT show the same numbers as our Power BI dashboards?

It reads the same models and measures your dashboards use, so the figures agree. When it filters differently from a report, it tells you which filters it applied.

### Can an AI agent refresh a Power BI dataset?

With read and write access and if your admins allow the tool, the agent can start a refresh of a semantic model and report its refresh history.