AI Tools Integration Examples

Learn how to connect your custom applications and data with AI models like ChatGPT, Claude, and your own LLMs.

🎯 What This Repo Covers

This repository demonstrates 4 different ways to integrate your data with AI systems:

Example	Description	Works With
mcp-server-stdio/	MCP Server (stdio transport)	Claude Desktop, Windsurf
mcp-server-sse/	MCP Server (SSE transport)	ChatGPT MCP Apps
chatgpt-actions/	REST API with OpenAPI	ChatGPT Custom GPTs
custom-llm-app/	Your own LLM + Tools	Databricks, OpenAI, any LLM

📚 Documentation

• AI-TOOLS-PRIMER.md - Complete guide explaining all concepts
• QUICK-START.md - Get running in 5 minutes

🚀 Quick Start

Prerequisites

# Python 3.9+
python --version

# Node.js (for MCP Inspector)
node --version

Install Dependencies

pip install -r requirements.txt

Run Examples

# MCP Server (stdio) - test with MCP Inspector
npx @modelcontextprotocol/inspector python mcp-server-stdio/server.py

# MCP Server (SSE) - run server, then connect Inspector
python mcp-server-sse/server.py
# Open http://localhost:8001

# ChatGPT Actions - REST API
python chatgpt-actions/main.py
# Open http://localhost:8000/docs

# Custom LLM App (requires Databricks credentials)
cp .env.example .env
# Edit .env with your credentials
python custom-llm-app/app.py
# Open http://localhost:8002

📁 Project Structure

ai-tools-example/
├── README.md
├── requirements.txt
├── .env.example
├── .gitignore
│
├── docs/
│   ├── AI-TOOLS-PRIMER.md      # Complete primer/guide
│   └── QUICK-START.md          # Quick start guide
│
├── mcp-server-stdio/           # MCP with stdio transport
│   ├── server.py
│   └── README.md
│
├── mcp-server-sse/             # MCP with SSE transport
│   ├── server.py
│   └── README.md
│
├── chatgpt-actions/            # REST API for ChatGPT
│   ├── main.py
│   └── README.md
│
├── custom-llm-app/             # Your own LLM app
│   ├── app.py
│   └── README.md
│
└── shared/                     # Shared data/utilities
    └── products.py

🔑 Key Concepts

How AI Tool Calling Works

User: "Show me pizzas"
         │
         ▼
┌─────────────────────────────┐
│  AI reads tool descriptions │
│  Decides: call get_products │
│  with category="pizza"      │
└─────────────────────────────┘
         │
         ▼
┌─────────────────────────────┐
│  Your server executes tool  │
│  Returns: [Pizza data...]   │
└─────────────────────────────┘
         │
         ▼
┌─────────────────────────────┐
│  AI formats response        │
│  "Here are our pizzas..."   │
└─────────────────────────────┘

Protocols Comparison

Protocol	Purpose	Use Case
MCP	AI ↔ Tools	Claude, Windsurf, ChatGPT
OpenAPI	API Documentation	ChatGPT Actions, REST clients
A2A	Agent ↔ Agent	Multi-agent systems

🛠️ Configuration

For Custom LLM App (Databricks)

Copy .env.example to .env and fill in your credentials:

DATABRICKS_HOST=https://your-workspace.cloud.databricks.com
DATABRICKS_TOKEN=your-token-here
DATABRICKS_ENDPOINT=databricks-claude-3-7-sonnet

For ChatGPT Actions (Public URL)

Use ngrok to expose your local server:

ngrok http 8000
# Use the https URL in ChatGPT

📖 Learn More

• MCP Documentation
• OpenAPI Specification
• LangChain Tools
• Databricks Model Serving

🤝 Contributing

Contributions welcome! Please read the contributing guidelines first.

📄 License

MIT License - see LICENSE file.