eatmundo MCP server

eatmundo is open to AI assistants through MCP (Model Context Protocol). Point your assistant at our endpoint and it can query 1,956 dishes from 26 cuisines in 7 languages — including the scored equivalence graph that connects a dish in one cuisine to its counterpart in another, with the shared ingredients as evidence. No account, no API key, no rate card.

What it answers that a model cannot

What is the Greek counterpart of lahmacun, and which ingredients do they actually share?
Build me a week of dinners with no pork and under 40 minutes each, one cuisine per day.
Compare köfte and keftedes ingredient by ingredient.

Install

The endpoint speaks streamable HTTP. Pick your client:

Claude Code

claude mcp add --transport http eatmundo https://www.eatmundo.com/api/mcp

Claude Desktop · Cursor · Windsurf (mcp.json)

{
  "mcpServers": {
    "eatmundo": {
      "type": "http",
      "url": "https://www.eatmundo.com/api/mcp"
    }
  }
}

MCP Inspector

npx @modelcontextprotocol/inspector https://www.eatmundo.com/api/mcp

Any MCP client works — these are just the common ones. The endpoint is the only thing you need.

Tools

find_similar_dishes: Dishes from other cuisines equivalent to a given one, from the scored graph. Every edge crosses country lines and is backed by shared canonical ingredients, not a guess.
compare_dishes: Two dishes ingredient by ingredient: shared, unique to each, a match score, and an editorial paragraph where one exists.
search_dishes: Search the corpus by name, cuisine, category or diet.
get_recipe: One full recipe: steps, FAQ, image credit, and ingredients with parsed quantities normalised to grams or millilitres.
weekly_menu: Up to 7 dinners honouring diet exclusions, a prep-time ceiling and a budget mode. When a constraint cannot be met it is relaxed and the relaxation is reported, never hidden.

Content is CC BY 4.0. Cite the URLs the tools return.

Listed in: SmitheryGlama