Menumigo for developers and AI agents

Quickstart

Three requests cover most integrations: find a restaurant, read its menu, read the plan limits. Authentication is not part of any of them.

1. Find a restaurant

curl -s "https://menumigo.com/api/public/restaurants?q=cafe&city=Lisbon&limit=3"

Each result carries the slug you need next, plus url, markdownUrl and apiUrl so you never have to build a URL by hand.

2. Read the full menu

curl -s https://menumigo.com/api/public/restaurants/portovibe-cafe

Returns categories and items with prices, currency, availability, dietary tags, the fourteen EU allergens, nutrition, variants, and translations keyed by language code. Serialization is an explicit allowlist, so no account or owner data is ever exposed.

3. Read the plan limits

curl -s https://menumigo.com/api/public/plans

A null limit means unlimited. Prices are in EUR.

Endpoint reference

The authoritative machine-readable version is openapi.json (OpenAPI 3.1). Everything below is a GET unless noted.

curl -s "https://menumigo.com/api/public"

curl -s "https://menumigo.com/api/public/restaurants?city=Lisbon&limit=5"

curl -s "https://menumigo.com/api/public/restaurants/portovibe-cafe"

curl -s "https://menumigo.com/api/public/locations"

curl -s "https://menumigo.com/api/public/cuisines"

curl -s "https://menumigo.com/api/public/plans"

curl -s "https://menumigo.com/api/public/tools/food-cost?cost=4.20&margin=70&price=14"

curl -s "https://menumigo.com/api/public/tools/wifi-qr?ssid=Cafe%20Guest&password=hello123"

curl -s "https://menumigo.com/api/public/tools/review-qr?placeId=ChIJN1t_tDeuEmsRUsoyG83frY4"

Error format

Every failure under /api is JSON — including a request to a path that does not exist. You never get an HTML error page from the API, so a parser can rely on the shape. Errors are sent with Cache-Control: no-store.

$ curl -s https://menumigo.com/api/public/restaurants/no-such-venue

{
  "error": "not_found",
  "status": 404,
  "message": "No public menu for slug \"no-such-venue\". Either no restaurant
              uses that slug, or its menu is not published.",
  "hint": "Search for the restaurant with
           https://menumigo.com/api/public/restaurants?q=no-such-venue and use the
           `slug` from the result.",
  "documentation": "https://menumigo.com/developers",
  "specification": "https://menumigo.com/openapi.json"
}

Branch on error — it is a stable code and will not be renamed. message describes what happened and hint describes the fix. A 404 on an unknown path also carries availableEndpoints, so a mistyped URL corrects itself in one round trip.

Markdown content negotiation

Every public page is served in two representations at one URL. Ask for Markdown with an Accept header, or append .md to the path — both return the same chrome-free document with YAML front matter and no navigation, scripts or styling.

curl -s -H "Accept: text/markdown" https://menumigo.com/pricing
curl -s https://menumigo.com/pricing.md          # same bytes, explicit URL
curl -s https://menumigo.com/index.md            # the home page

This follows the acceptmarkdown.com convention: responses carry Content-Type: text/markdown; charset=utf-8 and Vary: Accept, q-values are honoured (including q=0 as an explicit rejection), and a request that accepts none of text/html or text/markdown gets a 406 listing what is available rather than a silent fallback.

The complete list of Markdown pages is at /llms-index.md, generated from the route table rather than maintained by hand. Every HTML page also declares its twin with <link rel="alternate" type="text/markdown">, and every Markdown response points back with a Link: …; rel="canonical" header.

MCP server

Menumigo runs a remote Model Context Protocol server over HTTP JSON-RPC 2.0 at https://menumigo.com/api/mcp. It needs no credentials, so it can be added to any MCP client as a plain URL.

curl -s -X POST https://menumigo.com/api/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

initialize, tools/list, tools/call, resources/list and resources/read are supported. The manifest is at /.well-known/mcp.json.

Machine-readable manifests

Every discovery file this site publishes, in one place:

/openapi.json

OpenAPI 3.1 specification for every endpoint below.

/.well-known/mcp.json

MCP server manifest.

/.well-known/ai-plugin.json

AI plugin manifest.

/.well-known/agent.json

Agent discovery manifest: indexes, API map, Markdown pattern.

/.well-known/security.txt

Security contact and disclosure policy.

/llms.txt

Site overview written for agents.

/llms-full.txt

Full reference: every URL, feature and FAQ.

/llms-index.md

Every Markdown page, generated from the route table.

/sitemap.xml

Canonical HTML URLs.

Public restaurant menus additionally embed schema.org Restaurant + Menu JSON-LD, with hasMenuSection and hasMenuItem down to prices and allergens.

Limits, caching and terms

• No authentication, no quota registration. Please stay reasonable — cache responses and avoid hammering search. There is no published quota today; if throttling ever becomes necessary it will be announced here and returned as error: "rate_limited" with a 429, never as a silent failure.
• Cacheable by design. Directory and plan data carry a 24-hour s-maxage, restaurant search 60 seconds, and a single menu 5 minutes. Errors are never cached.
• Read-only. There is no write API. Menus are edited by their owners in the dashboard. Non-GET requests to a read endpoint return 405 with an Allow header.
• No personal data. The API exposes only what is already published on the public menu pages — no account data, no guest data, no analytics.
• Attribution. Data is provided for informational use with attribution to menumigo.com. Use is subject to our terms of service.

Something missing, or an endpoint you need? Tell us — the API is small on purpose and grows on request.