Skip to main content

AI Agents & MCP Server

Explore Córdoba runs a read-only Model Context Protocol (MCP) server. It exists so AI agents can query the site's monuments, events, restaurants, opening hours and walks directly instead of scraping rendered pages — see the Tools section below. The data is the site's own published content, refreshed hourly. The server needs no authentication and no API key. It only reads; it cannot write or change anything.

This page is a static entry point for search engines and agent crawlers. A page alone cannot make an agent discover or connect to the MCP server. An agent, or whoever operates it, still has to read the details below and configure the connection itself.

Connect

The discovery document above is separate from the _mcp.explorecordoba.com DNS TXT record some agent frameworks look up first. That TXT record only points to the discovery document; it is not the MCP endpoint itself.

Any MCP client that takes a JSON config

{
  "mcpServers": {
    "cordoba": { "type": "http", "url": "https://agents.explorecordoba.com/mcp" }
  }
}

Tools

  • events: List this city's festivals, fairs, processions and other events that overlap a date range (from and to, ISO dates; from defaults to today and to to 90 days after from), sorted by start date.
  • itinerary: Self-guided walking routes of this city from the site's own editorial walks.
  • monuments: Look up monuments, palaces, churches, gardens and other heritage sites of this city, either by exact slug or by free-text query over name, type, tags and description, optionally narrowed to a neighborhood.
  • opening_hours: Get the published opening hours of one monument or restaurant by slug (slugs come from the monuments and restaurants tools).
  • restaurants: Find restaurants and tapas bars of this city by neighborhood and/or a dish, cuisine or specialty (free text over cuisine, specialties, name and description; every word must match).

Before you rely on it

  • This is an early-access, experimental service meant to surface what agents actually ask for. It isn't a guaranteed data feed.
  • Requests are rate-limited per network address; the server returns an HTTP 429 with a Retry-After header when a client goes over.
  • Calls are logged for monitoring. IP addresses are never stored, and obvious personal data is stripped from logged arguments. Send no personal data in tool arguments regardless.

Questions about the site itself, rather than the MCP server?

Editorial standards Contact us