Power BI MCP Assistant

A natural language interface for Power BI, powered by Claude and the Model Context Protocol. Ask questions about your Power BI data in plain English — Claude handles the rest.

Available as both a browser-based chat UI (Streamlit) and a terminal REPL.

---

How it works

This app connects Claude to Microsoft’s Power BI MCP server. When you ask a question, Claude discovers the available Power BI tools, decides which ones to call, executes them, and returns an answer — often with an auto-rendered chart. The entire loop is agentic: Claude can chain multiple tool calls together to answer complex questions without you writing any DAX.

Features

• Natural language queries — no DAX required
• Agentic loop — Claude iteratively calls tools until it has a complete answer
• Auto-rendered charts — bar, line, pie, and metric cards generated from query results
• Two interfaces — Streamlit web UI and terminal CLI (REPL)
• Microsoft authentication — MSAL device code flow with local token caching
• Conversation memory — full multi-turn context within a session

Prerequisites

• Python 3.10+
• A Claude API key
• A Microsoft account with access to a Power BI workspace
• Access to Microsoft’s Power BI MCP server (https://api.fabric.microsoft.com/v1/mcp/powerbi)

Installation

git clone https://github.com/your-username/Connecting_PowerBI_MCP.git
cd Connecting_PowerBI_MCP

python -m venv .venv
source .venv/bin/activate      # Linux/Mac
# or
.venv\Scripts\activate         # Windows

pip install -r requirements.txt

Configuration

cp .env.example .env

Edit .env and set your values:

CLAUDE_API_KEY=sk-ant-...

# Optional — leave blank to use device code flow (recommended)
AZURE_TENANT_ID=
AZURE_CLIENT_ID=

That’s all you need for most setups. The app defaults to device code authentication, which means no Azure app registration is required — you just sign in through your browser on first run.

If you need a custom Azure app registration (e.g. for a service account or specific tenant), see docs/azure-setup.md.

Usage

Streamlit UI

streamlit run streamlit_app.py

Open http://localhost:8501 in your browser. On first run, you’ll be prompted to authenticate with Microsoft via device code flow.

CLI

python main.py

Available commands in the REPL:

Project structure

├── streamlit_app.py         # Streamlit entry point
├── main.py                  # CLI entry point
├── core/
│   └── orchestrator.py      # Agentic loop
├── clients/
│   ├── claude_client.py     # Anthropic API wrapper