Simple visuals + everyday analogies that explain MCP to everyone: whether you're a Power BI developer, a data engineer, or just curious how AI connects to real data.
If this helps you finally connect AI to your data, drop a ⭐. It helps more people find it.
AI models are powerful, but they're blind to your data. MCP (Model Context Protocol) is the open standard that fixes that: it lets AI read your databases, query your Power BI datasets, run notebooks in Fabric, and take actions in your data tools, safely and without custom glue code.
The problem: MCP docs are written for AI engineers, not data professionals. This repo sits in the middle. Every concept gets:
- 🧒 An "Explain Like I'm 5" analogy: the one-liner you'll actually remember
- 🖼️ A simple diagram: see how it connects, don't just read it
- 🔧 "How it actually works": for when you're ready to go deeper
- 🌍 A real-world example: where it actually helps in a data context
No AI engineering background required. No prior protocol experience needed. Just curiosity.
| # | Concept | One-liner |
|---|---|---|
| 1 | 🔌 What is MCP | The USB-C standard for connecting AI to tools and data. |
| 2 | 🏗️ MCP Architecture | Host, Client, Server: three roles, one protocol. |
| 3 | 🌐 MCP vs API | When the old way still works and when MCP is better. |
| # | Concept | One-liner |
|---|---|---|
| 4 | 📂 Resources | Data the AI can read: files, tables, live feeds. |
| 5 | 🛠️ Tools | Actions the AI can take: queries, writes, API calls. |
| 6 | 💬 Prompts | Reusable prompt templates the AI can discover and invoke. |
| 7 | 🌳 Roots | How the server tells AI which directories it can access. |
| 8 | 🎲 Sampling | Letting the MCP server ask the AI to generate something mid-task. |
| # | Concept | One-liner |
|---|---|---|
| 9 | 📊 MCP + Power BI | Let AI read your reports, datasets, and run DAX queries. |
| 10 | 🏭 MCP + Microsoft Fabric | AI on your lakehouse, warehouse, and pipelines. |
| 11 | 🗄️ MCP + SQL Databases | Query any database with natural language. |
| 12 | 📋 MCP + Excel | AI that can read and update your spreadsheets. |
| 13 | 🐍 MCP + Python / Pandas | Connect AI to your data science environment. |
| # | Concept | One-liner |
|---|---|---|
| 14 | 🔨 Building Your First MCP Server | The minimum viable server in under 50 lines. |
| 15 | 🔐 MCP Security | What to expose, what to protect, how auth works. |
| 16 | 🔍 MCP Inspector | The debugging tool every MCP developer needs. |
flowchart LR
U["👤 You / AI Agent"] --> H["🖥️ MCP Host\n(Claude, Cursor, VS Code)"]
H --> C["🔌 MCP Client"]
C --> S1["📊 Power BI\nMCP Server"]
C --> S2["🏭 Fabric\nMCP Server"]
C --> S3["🗄️ SQL\nMCP Server"]
S1 --> D1[("📈 Reports\n& Datasets")]
S2 --> D2[("🏔️ OneLake\n& Pipelines")]
S3 --> D3[("🗃️ Tables\n& Views")]
style H fill:#fef3c7,stroke:#f59e0b,color:#1f2937
style C fill:#dbeafe,stroke:#3b82f6,color:#1f2937
style S1 fill:#dcfce7,stroke:#22c55e,color:#1f2937
style S2 fill:#dcfce7,stroke:#22c55e,color:#1f2937
style S3 fill:#dcfce7,stroke:#22c55e,color:#1f2937
New to MCP? Start with What is MCP, then MCP Architecture. Then jump to whichever data tool you use most.
- Pick a concept from the table above.
- Read the analogy. Look at the diagram.
- Curious? Read "How it actually works."
- Found it useful? Star the repo ⭐ and share it.
Part of the Visual Edition series:
- 🏭 Microsoft Fabric (Visual Edition): OneLake, Lakehouses, DirectLake, and more
- 📊 DAX (Visual Edition): Filter context, CALCULATE, time intelligence
- 🗂️ Power BI Data Modeling (Visual Edition): Star schemas, relationships, performance
- 🧠 AI for Beginners (Visual Edition): LLMs, RAG, embeddings, and more
Know an MCP server we're missing? Have a better analogy? We'd love your help.
See CONTRIBUTING.md, adding a concept takes about 10 minutes.
Good first additions: MCP + Databricks, MCP + Snowflake, MCP + dbt, MCP + Azure Data Factory, MCP transport types (stdio vs SSE), MCP logging, MCP with LangChain, MCP with Semantic Kernel.
MIT: free to use, share, remix, and teach with. Attribution appreciated.
Made for data professionals who want AI that actually knows their data. 🔌
If this made MCP click for you, the best thank-you is a ⭐.