Cookwith

You can host the Cookwith MCP Server to generate and transform AI-powered recipes, adapting them to dietary needs, serving sizes, and nutrition targets. This MCP server exposes convenient tools for creating new recipes and modifying existing ones, enabling seamless integration with your culinary applications.

How to use

You interact with the Cookwith MCP Server through an MCP client. After configuring a connection, you can generate new recipes from natural language prompts and transform existing recipes to meet dietary goals or serving size changes. Use these tools to craft meals that align with calories, protein targets, and user preferences, then retrieve structured recipe data ready for display or further processing.

How to install

Prerequisites: you need Node.js and npm installed on your system. If you plan to run the server locally, ensure you have a compatible runtime environment for JavaScript.

Install via MCP Registry

npx @modelcontextprotocol/create-server install @cookwith/mcp-server

Install globally with npm

npm install -g @cookwith/mcp-server

Configure Claude Desktop (example for macOS and Windows). This integrates Cookwith MCP Server into your Claude Desktop setup.

{
  "mcpServers": {
    "cookwith": {
      "command": "npx",
      "args": ["@cookwith/mcp-server"]
    }
  }
}

Additional setup notes

If you run in a local development environment, you can start the server after installation using the standard start command defined by the project. Typical workflow involves cloning the repository, installing dependencies, building, and then starting the server. Ensure you can access the MCP API endpoint for client integrations.

Configuration and endpoints

The server exposes an HTTP API that you can query for recipe generation and transformation. Use the production endpoint for live usage and the development endpoint for local testing.

Production: https://cookwith.co/api/mcp
Development: http://localhost:3000/api/mcp

Example usage flow

1. Connect your MCP client to the Cookwith MCP Server using the provided HTTP URL or the local stdio command. 2. Generate a recipe by supplying a natural language prompt along with optional dietary restrictions, allergies, and serving targets. 3. Transform an existing recipe by providing the recipe object, transformation instructions, and any new nutritional targets or serving size.

Tool endpoints and capabilities

The server provides two primary tools you can call through your MCP client.

generate_recipe

Create a new recipe from a natural language description while optionally applying dietary restrictions, allergies, dislikes, calorie targets, and a specified number of servings.

transform_recipe

Modify an existing recipe by applying transformation instructions and updating calories, protein, or servings as needed.

Available tools

generate_recipe

Generate a new recipe based on a natural language prompt while applying dietary restrictions, allergies, calories, protein targets, and servings.

transform_recipe

Transform or modify an existing recipe according to given instructions, with optional new calories, protein, or serving targets.