Content
# 🍳 HowToCook-MCP Server 🥘 -- Cook delicious meals for a week, say no to mediocre food
[English](./README_EN.md) | 简体中文
<div align="center">
This project is sponsored by Tencent EdgeOne for CDN acceleration and security protection.
[Asia's Best CDN, Edge, and Security Solutions - Tencent EdgeOne](https://edgeone.ai/zh?from=github)
<img src="./public/edgeone.png"/>
</div>
> Let the AI assistant transform into your personal chef, helping you plan your meals for the day!
Based on [Anduin2017/HowToCook](https://github.com/Anduin2017/HowToCook), the MCP (Model Context Protocol) server allows the AI assistant to recommend recipes and plan meals, solving the century-old question of "What to eat today?"
Data source: [Anduin2017/HowToCook](https://github.com/Anduin2017/HowToCook) ⭐ If you haven't starred it yet, go give it a star!
🎉 **Want to use the current MCP directly? Experience it now at** [https://howtocookmcp.weilei.site/](https://howtocookmcp.weilei.site/)
🎉 **Additionally, we also provide DXT (Desktop Extensions) for everyone to experience, easily installable to Claude Desktop.**
As follows: Please ensure you have the latest version of Claude Desktop installed. The current MCP's DXT file has been uploaded to the repository, and you can download it or fork this repository to build it yourself.
How to package for local development into DXT?
1. Run `npm install -g @anthropic-ai/dxt`
2. In the folder containing the local MCP server, run `dxt init`. This is the root directory of your MCP, and this command will guide you to create `manifest.json`.
3. Run `dxt pack` to create the dxt file.
Now, any application that supports DXT can run your local MCP server. For example, open the file with Claude for macOS and Windows to display the installation dialog.
Refer to: [anthropics/dxt](https://github.com/anthropics/dxt)
## 📸 Preview
## 🔌 Supported MCP Clients
This server is compatible with all AI assistants and clients that support the MCP protocol, including but not limited to:
- 🤖 Claude Desktop Application
- 📝 Cursor
- 💼 Other clients that support MCP
## ✨ Delicious Features
This MCP server provides the following culinary tools:
1. **📚 Query All Recipes** - Get all available recipe data, a cooking encyclopedia -- use with caution -- context is too large.
2. **🔍 Query Recipes by Category** - Filter recipes by category. Want seafood? Breakfast? Meat dishes? Staples? One-click solution!
3. **📖 Query Specific Recipe** - Query the complete details of a specific recipe by name, including ingredients, steps, etc.
4. **🧩 Smart Meal Recommendations** - Plan a whole week of delicious dishes based on your dietary restrictions, allergens, and number of diners.
5. **🎲 Not sure what to eat?** - A blessing for those with indecision! Directly recommend today's menu based on the number of people, no more dilemmas.
## 🚀 Quick Start
### 📋 Prerequisites
- Node.js 16.0.0+ 🟢
- npm or yarn 📦
### 💻 Installation Steps
1. Clone the culinary repository
```bash
git clone https://github.com/worryzyy/howtocook-mcp.git
cd howtocook-mcp
```
2. Install dependencies (as simple as preparing ingredients!)
```bash
npm install
```
3. Compile the code (the cooking process...)
```bash
npm run build
```
### 🎯 Command Line Parameters
The server supports the following command line parameters:
- `--transport <stdio|http|sse>` - Choose the transport method (default is stdio)
- `--port <number>` - Listening port when using http or sse transport (default is 3000)
Example: Use http transport and listen on port 8080
```bash
node build/index.js --transport http --port 8080
```
## 🍽️ Getting Started
### 🔥 Start the Server
```bash
npm start
```
### 🔧 Configure MCP Client
#### Recommended to use Cursor for a quick experience (two methods)
1. Using npm package: First run `npm i -g howtocook-mcp`, otherwise you will encounter `Failed to create client`.
Then add the MCP server configuration in Cursor settings:
```json
{
"mcpServers": {
"howtocook-mcp": {
"command": "npx",
"args": ["-y", "howtocook-mcp"]
}
}
}
```
2. If running locally from a cloned repository, please use the following configuration:
```json
{
"mcpServers": {
"howtocook-mcp": {
"command": "node",
"args": ["yourpath\\howtocook-mcp\\build\\index.js"]
}
}
}
```
#### Other MCP Clients
For other clients that support the MCP protocol, please refer to their respective documentation for configuration, usually requiring:
- Server Name: `howtocook-mcp`
- Command: `npx -y howtocook-mcp`
3. Restart the client to let the culinary magic take effect ✨
## 🧙♂️ Menu Magic Usage Guide
Here are example prompts used in various MCP clients:
### 1. 📚 Query All Recipes
No parameters needed, just summon the culinary encyclopedia!
```
Please use the howtocook MCP service to query all recipes.
```
### 2. 🔍 Query Recipes by Category
```
Please use the howtocook MCP service to query seafood recipes.
```
Parameters:
- `category`: Recipe category (seafood, breakfast, meat dishes, staples, etc.)
### 3. 🧩 Smart Weekly Recipe Recommendation
```
Please use the howtocook MCP service to recommend a week's worth of recipes for 3 people. We don't eat cilantro and are allergic to shrimp.
```
Parameters:
- `allergies`: List of allergens, e.g., ["garlic", "shrimp"]
- `avoidItems`: Ingredients to avoid, e.g., ["scallions", "ginger"]
- `peopleCount`: Number of diners (1-10)
### 4. 🎲 What to eat today?
```
Please use the howtocook MCP service to recommend a dinner menu for 4 people.
```
Parameters:
- `peopleCount`: Number of diners (1-10)
## 📝 Tips
- This package has been published to npm and can be globally installed via `npm install -g howtocook-mcp`.
- This service is compatible with all AI assistants and applications that support the MCP protocol.
- On first use, the AI may need some time to familiarize itself with how to use these tools (just like heating up a pan).
## 🤝 Contribution
Feel free to fork and submit pull requests to help us improve this culinary assistant!
## 📄 License
MIT License - Use freely, as generously as sharing a recipe!
---
> 🍴 The culinary adventure is about to begin, are you ready to satisfy your appetite?
