MCP server for AI agents

Server URL

Streamable HTTP in stateless mode. Any MCP-compatible client can add it with this URL:

https://www.beachvalencia.com/api/mcp

Available tools

All four tools are read-only (readOnlyHint: true). None of them mutates data, charges money or books anything.

• search_availability — availability search for a date range and party size across every category.
• get_property_details — description, capacity, photos and generic location for one category (G1..G6, G4B).
• get_quote — exact quote for a specific category, including deposit and cancellation summary.
• get_policies — check-in / check-out times, cancellation, deposit, Spanish traveller registration (SES.Hospedajes) and contact channels.

Example prompts

• “Find me a two-bedroom apartment near the beach in Valencia for four people, 15 to 20 October, under 900 €.”
• “What’s the exact price for the largest apartment with pool from 24 December to 1 January for six people?”
• “What’s the cancellation policy if I book now and cancel one week before arrival?”

What it does NOT do

• It does not book or charge from the chat: any booking and payment are completed on beachvalencia.com with the same checkout as a normal direct booking.
• It does not expose apartment codes (Axxx-N) or exact street addresses. The specific address and access instructions are sent by email once the reservation is confirmed.
• It does not validate or apply discount codes. If you have one, you enter it yourself on the website checkout.
• It does not access personal guest data or booking history.
• Prices returned are the public non-member rate with no discount code — exactly what you would pay at the website checkout.

Data we log

Each call to the MCP server is logged anonymously for operational purposes, capacity planning and abuse detection: timestamp, tool invoked, query parameters (dates, party size, category), latency, result and — if the MCP client sends it — the client identifier itself (e.g. "claude.ai" or "chatgpt.com"). We do not store your IP address in clear text: it is turned into an irreversible hash (SHA-256 of IP + user-agent + an internal secret) used only to rate-limit sessions.

Limits and kill switch

• Up to 600 calls every 10 minutes per anonymous session (ip_hash).
• Global upstream cap to our booking engine: 120 calls per minute.
• Bursts above the limit return a readable error with retry_after_seconds; the server does not retry aggressively.
• Operational kill switch (MCP_ENABLED=false) returns HTTP 503 with a readable message without hitting the booking engine.

Public registry

We are listed in the official Model Context Protocol registry under the canonical name com.beachvalencia/booking. Namespace ownership is verified via DNS (Ed25519).

• Listing: https://registry.modelcontextprotocol.io/v0/servers?search=com.beachvalencia/booking

Legal and contact

Use of the server is subject to our Privacy Policy and Rental terms and conditions.

• Privacy: https://www.beachvalencia.com/privacidad
• Terms and conditions: https://www.beachvalencia.com/legal/condiciones
• Operational contact (bugs, incidents, feedback): reservas@beachvalencia.com