HeatherFlux/MCP-Demo-Recipes
If you are the rightful owner of MCP-Demo-Recipes and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
This document provides a structured summary of a Model Context Protocol (MCP) server designed to demonstrate AI's ability to chain multiple tools for solving complex user queries.
Recipe Assistant MCP Server Demo
A Model Context Protocol (MCP) server that demonstrates how AI can intelligently chain multiple tools to solve complex user problems.
What This Demonstrates
This server showcases multi-tool AI workflows where an AI assistant can:
- Understand a natural language request
- Determine which tools to use and in what order
- Pass data between tools automatically
- Provide a complete solution from a single user query
The Demo Scenario
User says: "I want to make seafood paella for 6 people tonight"
AI automatically:
- 🔍 Searches for paella recipes
- 📋 Gets the full recipe details
- 🔢 Scales ingredients from 4 to 6 servings
- 📏 Converts to preferred measurement system (optional)
- 🛒 Organizes into a shopping list by grocery section
- 💰 Estimates the total cost
Result: User gets a complete shopping list with cost estimate, ready to shop!
Available Tools
| Tool | Purpose |
|---|---|
search_recipes | Search TheMealDB API for recipes |
get_recipe_details | Get full recipe with ingredients |
scale_recipe | Adjust serving sizes |
convert_units | Metric ↔ Imperial conversion |
create_shopping_list | Organize by store section |
estimate_cost | Budget calculation |
Setup & Running
# Install dependencies
pnpm install
# Build and run the server
pnpm run dev
# Or build and start separately
pnpm run build
pnpm start
Server runs on http://localhost:3000/mcp
Using MCP Inspector for Testing
In one terminal, start the server:
pnpm start
In another terminal, start the inspector:
pnpm run inspect
Then in the inspector UI:
- Select "HTTP (SSE)" transport
- Enter URL:
http://localhost:3000/mcp - Click Connect
Now you can test all the tools interactively!
How It Works
This server uses:
- MCP SDK for protocol handling
- StreamableHTTPServerTransport for HTTP-based tool calls
- TheMealDB API for real recipe data (no API key needed!)
- Zod for schema validation
Each tool is registered with input/output schemas so AI clients know how to use them and can chain them intelligently.
Testing the Demo
Connect an MCP-compatible AI client (like Claude Desktop or your own client) to http://localhost:3000/mcp and try:
"I want to make chicken tikka masala for 8 people"
"Find me a recipe for beef stroganoff, scale it for 3 people, and show me what it will cost"
"I'm cooking paella for my dinner party of 6 tonight, help me plan"
The AI will automatically determine which tools to call and in what sequence!
Why This Matters
Traditional chatbots can only respond with text. With MCP:
- AI can take actions by calling tools
- Tools can be chained together for complex workflows
- One simple request → complete solution
- Demonstrates practical AI utility beyond conversation
Architecture Highlights
- Each HTTP request gets a new transport instance (prevents ID collisions)
- Tools return both text content and structured data
- Schemas guide AI on proper tool usage
- Real external API integration (TheMealDB)
See CLAUDE.md for detailed implementation notes.