Summary
It produces the shopping list as part of the plan, which is the half people usually do by hand.
mcp_howtocook_recommendMeals takes headcount, allergies and the things you will not eat, and returns both the week's plan and the list you shop from. The rest of the surface answers the follow-up question: pick a category, then pull the full ingredients and steps for the dish you settled on.
What it is
A local server over the HowToCook recipe collection, five tools: list every recipe, filter by one of ten categories, fetch one recipe in full, get a dish combination for a headcount, and build a week's meal plan with a shopping list.
What you get
- Browse the whole recipe set (mcp_howtocook_getAllRecipes), or filter to one of ten categories — seafood, breakfast, seasonings, desserts, drinks, meat dishes, prepared foods, soups, staples and vegetable dishes (mcp_howtocook_getRecipesByCategory).
- Fetch one recipe in full, ingredients and steps, by name or ID (mcp_howtocook_getRecipeById).
- Get a dish combination sized to the number of people eating (mcp_howtocook_whatToEat).
- Build a week's meal plan and its shopping list, with allergens and disliked ingredients excluded (mcp_howtocook_recommendMeals).
Requirements
Node.js on the machine; no account and no key. MIT licensed. The collection is Chinese-language, category names included.
Setup effort
One command — npx -y howtocook-mcp