Mei Recipes MCP

Use the Model Context Protocol to let AI assistants read, search, and manage your recipes directly.

Setup

Connect to the Mei Recipes MCP server over the internet — no cloning or local setup required. Your AI client connects directly to the hosted API.

1. Add to your MCP client

Add the following to your MCP client configuration.

json

{
  "mcpServers": {
    "mei-recipes": {
      "url": "https://mei-recipes-api.karthik-rk.workers.dev/mcp"
    }
  }
}

Claude Code

Add the config to .mcp.json in any project directory — Claude Code will auto-detect it.

Claude Desktop

Go to Settings → Developer → Edit Config and add the server entry to your claude_desktop_config.json.

Cursor / Windsurf

Add the configuration to your editor's MCP settings.

Authorisation

Reading recipes is public — no token needed. To add or modify recipes, your MCP client must send a service token via the Authorization header. Ask the Mei Recipes admin for a token.

How it works

Your AI client connects to the remote MCP endpoint over HTTPS. The server exposes the same tools as the local version — all reads, writes, and searches work identically. No data is stored on your machine.

Available Tools

9 tools are available for AI agents to interact with your recipe collection.

list_recipes

List all recipes with summaries. Optional text search by title or tag.

get_recipe

Get a full recipe by ID with ingredients, steps, nutrition, and metadata.

list_wip

List all WIP items with summaries. Optional search by note, tags, or description.

get_wip

Get a full WIP item by ID with all fields.

search_recipes

Search recipes by ingredient name, tag, or cuisine.

add_recipe

Add or update a fully structured recipe with ingredients, steps, and nutrition.

add_wip

Add a new WIP item from a URL, name, or description.

convert_wip

Convert a WIP item into a full recipe. Saves the recipe and removes the WIP entry.

delete_wip

Delete a WIP item by ID.

Schema Resource

The server exposes a mei://schema/recipe resource containing type definitions, a JSON schema reference, and a sample recipe. AI agents should read this before creating or converting recipes.

Example Prompts

Try these with any MCP-connected AI assistant.

Search and compare

“What chicken recipes do I have? Show me the ones with the shortest cook time.”

Quick save

“Add a WIP for this recipe: https://www.bonappetit.com/recipe/crispy-chicken”

WIP to recipe

“Convert the "spicy-tofu-bowl" WIP into a full recipe with ingredients and step-by-step instructions.”

Notes

•

Ingredient IDs are 4-character zero-padded strings (e.g. “0001”). Steps reference them with {ingredient_id} syntax.

•

When converting WIP items, the AI agent should read the schema resource first, then the WIP item, then call convert_wip.

•

Read operations don't require authorisation. Write operations need a service token.