HowToCook

HowToCook-MCP Server lets AI assistants recommend recipes, plan meals, and answer  to the everyday question of what to eat. It exposes a structured MCP (Model Context Protocol) interface that you can connect to from any MCP-capable client to explore recipes, filter by category, get full recipe details, and generate smart weekly menus tailored to your preferences and constraints.

How to use

Connect with any MCP-enabled client to access the HowToCook MCP Server. You can browse the full recipe collection, search by category, pull details for a specific dish, receive a smart weekly meal plan, or get a suggested menu for today based on the number of people and your dietary restrictions.

How to install

Prerequisites: install Node.js 16.0.0+ and a package manager such as npm or yarn.

Step 1: Clone the project repository and navigate into it.

git clone https://github.com/worryzyy/howtocook-mcp.git
cd howtocook-mcp

Step 2: Install dependencies.

npm install

Step 3: Build the server code.

npm run build

Step 4: Start the MCP server (HTTP transport on port 8080 for example). The final runtime command starts the server with HTTP transport.

node build/index.js --transport http --port 8080

Additional configuration and usage notes

Two ready-to-use MCP connection options are available. The HTTP endpoint provides a remote MCP URL, while the local runtime can be started directly on your machine.

// HTTP connection (remote MCP URL)
{
  "type": "http",
  "name": "howtocook_mcp",
  "url": "https://howtocookmcp.weilei.site/mcp",
  "args": []
}

Security and deployment tips

If you plan to expose the MCP server publicly, consider enabling transport security and applying access controls at the network level. Keep dependencies up to date and monitor for any security advisories related to the server stack.

Troubleshooting

If clients fail to connect, verify that the server is running on the expected port and that the MCP client is configured with the correct server name and transport. Check console logs for any startup errors and ensure that the build output exists at build/index.js.

Notes

The server supports multiple MCP clients and can be run locally or accessed via a remote MCP URL. First-time use may require a moment for the client to establish context and familiarize with available tools.

Available tools

queryAllRecipes

Fetches all available recipes from the data set, serving as a comprehensive reference similar to a recipe encyclopedia.

queryByCategory

Filters recipes by category such as seafood, breakfast, main dishes, or staples.

queryRecipe

Retrieves complete details for a specific recipe, including ingredients and step-by-step instructions.

recommendWeeklyMenu

Generates a personalized week-long menu based on allergies, dietary preferences, and number of people.

randomTodayMenu

Suggests a menu for today tailored to the group size and restrictions.