Powerbi Claude Mcp Pipeline

An end-to-end Business Intelligence pipeline that integrates Claude AI via Model Context Protocol (MCP) into a real Power BI workflow — from raw data cleaning to AI-assisted dashboard development.

---

Dashboard Preview

---

Overview

This project demonstrates a complete, production-style BI pipeline where Claude AI acts as an active collaborator — not just a chatbot — helping build DAX measures, answer business questions, validate data, and create visuals directly within Power BI.

---

Pipeline — Step by Step

1. Data Cleaning (Python + Pandas)

• Ingested raw data and performed cleaning using Python and Pandas
• Handled null values, standardised field names, and validated record integrity
• Ensured data was analysis-ready before loading to the database

2. Database Layer (MS SQL Server)

• Established a Python-to-SQL Server connection using pyodbc / sqlalchemy
• Pushed the cleaned dataset into a structured SQL Server table
• Created a reliable, queryable single source of truth for downstream reporting

3. Power BI Connection

• Connected Power BI directly to the SQL Server table
• Loaded the clean dataset into the data model for reporting and visualisation

4. Claude MCP Integration

• Downloaded and configured the Claude MCP server
• Established a live AI-to-Power BI bridge
• Enabled Claude to interact directly with the Power BI data model

5. AI-Assisted Analysis & Development

With the MCP integration active, Claude was used to:

• Write and validate DAX measures for business KPIs
• Answer business questions in natural language (counts, aggregations, trends)
• Audit data quality — verifying total record counts and identifying anomalies
• Rename and standardise field labels for business-friendly reporting
• Recommend and generate appropriate visuals for each metric

---

Features

• AI-assisted Power BI development using Claude MCP
• Natural language querying against BI datasets
• Automated DAX measure generation
• SQL Server integration pipeline
• Python-based data cleaning and validation
• AI-assisted visual recommendation workflow
• Production-style BI architecture

Example AI Queries

With Claude MCP connected to Power BI, the assistant can respond to prompts such as:

- "Create a DAX measure for YoY sales growth"
- "Which customer segment generated the highest revenue?"
- "Recommend the best visual for monthly sales trends"
- "Validate total order counts against the SQL source table"
- "Show the top 5 products by profit margin"

Tech Stack

Layer	Technology
Data cleaning	Python 3.x, Pandas
Database	Microsoft SQL Server
BI & Visualisation	Power BI
AI Integration	Claude AI, MCP Server
Query & Measures	DAX, SQL

---

Architecture Flow

Raw CSV Data
    ↓
Python + Pandas Cleaning
    ↓
SQL Server Database
    ↓
Power BI Data Model
    ↓
Claude MCP Integration
    ↓
AI-Assisted Analytics & DAX Generation

Project Structure

powerbi-claude-mcp-pipeline/
├── README.md
├── data/
│   └── sample_data.csv          # Anonymised sample dataset
├── python/
│   ├── data_cleaning.py         # Data cleaning and transformation
│   └── sql_upload.py            # SQL Server connection and data upload
├── powerbi/
│   └── dashboard_screenshot.png # Final dashboard preview
├── mcp/
│   └── mcp_config.json          # MCP server config (credentials removed)
├── .env.example                 # Environment variable template
├── requirements.txt             # Python dependencies
└── .gitignore

---

Getting Started

Prerequisites

• Python 3.x
• Microsoft SQL Server (local or remote)
• Power BI Desktop
• Anthropic API key (get one here)
• Claude MCP server

Installation

# Clone the repository
git clone https://github.com/aakshadha/powerbi-claude-mcp-pipeline.git
cd powerbi-claude-mcp-pipeline

# Install Python dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env
# Edit .env with your SQL Server credentials and Anthropic API key

Running the Pipeline

# Step 1: Clean the data
python python/data_cleaning.py

# Step 2: Upload to SQL Server
python python/sql_upload.py

# Step 3: Open Power BI Desktop and connect to your SQL Server table
# Step 4: Configure the MCP server (see mcp/mcp_config.json)
# Step 5: Connect Claude and start querying your data

---

Key Learnings

• MCP enables Claude to act as a BI collaborator — it understands the data model and contributes directly to report development, not just answers generic questions
• Combining a clean data pipeline with AI assistance dramatically reduces the time from raw data to business insight
• DAX measure creation, which typically requires specialist knowledge, can be accelerated significantly with Claude's assistance

---

Security Notes

• Never commit real credentials — use .env files and environment variables
• The mcp_config.json file in this repository contains placeholder API keys only
• Sample data has been anonymized — do not upload real business or personal data

---

Author

Anjali Akshadha Singh
BI Engineer | Power BI · Snowflake · SQL | AI-augmented analytics
LinkedIn

---

License

MIT License — feel free to use and adapt this project.