Hotel & Vacation Rental MCP - live prices, availability & reviews across Booking.com, Airbnb, Vrbo & Google Hotels

Search stays, compare prices across Booking.com, Airbnb, Vrbo and Google Hotels, and pull reviews by just asking your AI. One hosted MCP, 300 free credits to start, no card.
Seven read-only tools - search, availability, listing detail, price, cross-OTA price compare, reviews, and job polling - for Claude, ChatGPT, Cursor, Claude Code, and any MCP client.

4 booking platforms, one schema · 7 read-only tools · cross-OTA price compare · OAuth 2.1, no key pasted into the agent · 300 free credits to start, no card StayingAPI is an independent service for accommodation data. StayingAPI is not affiliated with, endorsed by, or sponsored by Airbnb, Booking.com, Vrbo, or Google Hotels. These are trademarks of their respective owners, used here descriptively to indicate the data sources StayingAPI's API and MCP server can query.

Install in one line

Claude Code

claude mcp add --transport http stayingapi https://mcp.stayingapi.com/mcp

Claude Desktop, ChatGPT, Cursor, or any other MCP client — add this as a Streamable HTTP server:

https://mcp.stayingapi.com/mcp

No API key to paste: the first call opens an OAuth sign-in (free account, 300 credits, no card). Per-client steps are in Quick Install, and the whole repo also installs as an Agent Plugin.

Why a Hotel & Vacation Rental MCP

You are already comparing stays across four sites by hand: a hotel on Booking.com, an apartment on Airbnb, a villa on Vrbo, and Google Hotels to sanity-check the rate. Every one of them has a different page, a different rating scale, and no API you can just sign up for.

Connect once, then just ask. Add this MCP server to Claude, ChatGPT, Cursor, or any MCP client a single time, and your AI assistant can search real stays, quote real prices for your dates, compare one property's rate across booking sites, and read normalized reviews - inside the conversation, with no code and no repeated setup.

Everything comes back in one unified schema. A hotel room and a holiday villa return the same object shape, ratings carry their native scale instead of being silently rescaled, and the cross-OTA comparison returns each site's offer plus a computed lowest and median price, so you do not re-derive them yourself.

Quick taste:

Find 2-bed stays in Lisbon for these dates under EUR 150,
compare the top one's price across Airbnb, Booking.com and Vrbo,
and summarize its reviews.

That single prompt spans three tools - search_stays, compare_prices, get_reviews - without you writing a line of code.

Install as an Agent Plugin _{· easiest}

This repository root is a conformant Agent Plugins 1.0.0 package — the portable format published 2026-08-06 and supported by ChatGPT, Codex, Cursor, GitHub Copilot, Kiro and VS Code. One install gets you the MCP server and a bundled stays skill that teaches your agent which tool answers which question, how to keep cross-platform fan-out cheap, and that Airbnb/Vrbo rate on a 5-point scale while Booking.com rates on 10 — so it never compares a 9 with a 4.5.

plugin.json                    # Agent Plugins 1.0.0 manifest
mcp.json                       # hosted MCP server, streamable-http, OAuth (no keys)
skills/stays/SKILL.md          # when + how to use the 7 read-only tools
.cursor-plugin/plugin.json     # Cursor plugin manifest, same MCP + skill (+ marketplace.json)

VS Code — Command Palette → Chat: Install Plugin From Source, then paste:

https://github.com/stayingapi/hotel-vacation-rental-mcp

Or register a local clone in settings.json:

"chat.pluginLocations": { "/absolute/path/to/hotel-vacation-rental-mcp": true }

Cursor — Customize in the sidebar → find the plugin → Install. For a local clone:

git clone https://github.com/stayingapi/hotel-vacation-rental-mcp ~/.cursor/plugins/local/stayingapi

Then Developer: Reload Window.

Add to Cursor — MCP server only, no clone. Paste this deeplink into your browser:

cursor://anysphere.cursor-deeplink/mcp/install?name=stayingapi&config=eyJ1cmwiOiJodHRwczovL21jcC5zdGF5aW5nYXBpLmNvbS9tY3AifQ==

ChatGPT, Codex, GitHub Copilot, Kiro, any other client — point your client's plugin mechanism at this repository, or at a local clone. Agent Plugins 1.0.0 standardizes the package format, not installation, so each client owns its own install flow.

30-second example

Find a 2-bedroom apartment in Split, Croatia for 12-15 June, 2 adults, under EUR 150 a night.

One search_stays call returns Airbnb, Booking.com, Vrbo and Google Hotels merged into a single normalized list, each with its booking URL. No API key to configure — the first call opens an OAuth sign-in (free account, 300 credits, no card). Then the follow-ups that normally take twenty browser tabs:

Is that one cheaper on Booking or Airbnb?   → compare_prices (one call, every OTA)
Is it free all week, and what's the min stay? → check_availability (day by day)

All seven tools are read-only — nothing here can book, cancel or change a reservation.

There are no credentials in this package — Agent Plugins 1.0.0 forbids embedded secrets, and authorization is client-managed. Verify the package yourself:

curl -sO https://agent-plugins.org/schemas/1.0.0/plugin.schema.json
curl -sO https://agent-plugins.org/schemas/1.0.0/mcp.schema.json
npx ajv-cli@5 validate --spec=draft2020 -s plugin.schema.json -d plugin.json
npx ajv-cli@5 validate --spec=draft2020 -s mcp.schema.json    -d mcp.json

Quick Install

Requirements:

• A StayingAPI account (sign up - 300 free credits, no card)
• Either OAuth (Claude, ChatGPT - no key handling at all) or an API key (stay_live_... / stay_test_...) from your dashboard

Recommended: add a rule so your AI invokes it automatically

Paste this into your client's custom instructions or rules file:

txt

Copy

When I mention a hotel, an Airbnb/Booking.com/Vrbo/Google Hotels listing,
a place to stay, travel dates, or ask to compare accommodation prices or
read stay reviews, automatically use the StayingAPI MCP tools to fetch
real data before answering.

Server URL: https://mcp.stayingapi.com/mcp (Streamable HTTP transport)

https://mcp.stayingapi.com/mcp

{
  "mcpServers": {
    "stayingapi": {
      "url": "https://mcp.stayingapi.com/mcp"
    }
  }
}

claude mcp add --transport http stayingapi https://mcp.stayingapi.com/mcp

claude mcp add --transport http stayingapi https://mcp.stayingapi.com/mcp \
  --header "Authorization: Bearer stay_live_YOUR_KEY_HERE"

https://mcp.stayingapi.com/mcp

{
  "mcpServers": {
    "stayingapi": {
      "url": "https://mcp.stayingapi.com/mcp"
    }
  }
}

Authentication

Two ways in, both ending at the same account and the same credit balance:

• OAuth 2.1 + PKCE (S256) with Dynamic Client Registration - recommended for Claude and ChatGPT. You authorize once in a browser popup and the connector is linked to your StayingAPI account. DCR-aware clients discover everything automatically, so no key is ever pasted into the agent.
• Bearer API key - for key-based clients and scripts. Keys are stay_live_... (real fan-out to live sources) or stay_test_... (deterministic sandbox fixtures, always 0 credits). Send as Authorization: Bearer stay_live_....

One balance, separate rate bucket. MCP calls draw from the same single credit balance as your REST keys. REST is rate-limited per key; MCP is rate-limited per user - separate buckets, one wallet.

Security note: treat stay_live_ keys like passwords. A new key's secret is shown once and stored only as a SHA-256 hash, so it is never recoverable. Never commit one to a repo or paste it into a shared config; revoke and re-mint from the dashboard if a key leaks.

Available Tools

All seven are read-only (annotated readOnlyHint: true, idempotentHint: true). There are no write tools, so nothing here can book, cancel, or change anything. Each tool maps 1:1 to a REST endpoint, runs the same validation and cache pipeline, and returns the same unified { data, meta } envelope.

1. search_stays

Discover properties across platforms by location, dates, occupancy and filters, merged into one schema. This is the breadth endpoint and the clearest demonstration of "one schema, every platform".

Response (trimmed):

{
  "data": [
    {
      "id": "stays_booking_abramovic2",
      "platform": "booking",
      "platformListingId": "abramovic2",
      "url": "https://www.booking.com/hotel/hr/abramovic2.html",
      "name": "Apartments Abramović",
      "propertyType": "apartment",
      "location": { "lat": 43.51, "lng": 16.44, "city": "Split", "country": "HR" },
      "starRating": null,
      "guestRating": 9.1, "ratingScale": 10, "reviewCount": 142,
      "maxOccupancy": 4, "bedrooms": 2, "bathrooms": 1,
      "amenities": ["pool", "kitchen", "air_conditioning", "wifi"],
      "host": { "name": "Marko", "isSuperhost": false },
      "price": { "currency": "USD", "nightlyPrice": 303, "totalPrice": 2122, "nights": 7 }
    }
  ],
  "meta": {
    "platforms": ["airbnb", "booking"],
    "cached": false, "partial": false, "currency": "USD",
    "pagination": { "limit": 20, "cursor": null, "nextCursor": "eyJ…", "hasMore": true },
    "platformResults": [
      { "platform": "airbnb", "status": "ok", "cached": false, "count": 18 },
      { "platform": "booking", "status": "ok", "cached": true, "count": 20 }
    ],
    "warnings": []
  }
}

2. check_availability

Day-by-day availability for a known listing (or a batch of listings) on one platform over a date window. Each day reports whether it is available, its minimum-night requirement, and whether check-in, check-out and booking are allowed.

Response (trimmed):

{
  "data": [
    {
      "platform": "airbnb",
      "listingId": "42307961",
      "dates": [
        { "date": "2026-07-13", "available": true, "minNights": 7, "checkIn": true, "checkOut": false, "bookable": true },
        { "date": "2026-07-14", "available": true, "minNights": 7, "checkIn": false, "checkOut": false, "bookable": true }
      ]
    }
  ],
  "meta": { "platforms": ["airbnb"], "cached": false, "partial": false, "warnings": [] }
}

3. get_listing

Full normalized detail for one listing: amenities on a canonical taxonomy, photos, host, geo, ratings, and - when you pass dates - an embedded live price. Detail is cached 24 h and any embedded price at the 1 h price TTL, composed at read time, so a stale price never rides on fresh detail.

Response (trimmed):

{
  "data": {
    "id": "stays_booking_abramovic2",
    "platform": "booking",
    "platformListingId": "abramovic2",
    "name": "Apartments Abramović",
    "propertyType": "apartment",
    "location": { "lat": 43.51, "lng": 16.44, "city": "Split", "country": "HR" },
    "guestRating": 9.1, "ratingScale": 10, "reviewCount": 142,
    "maxOccupancy": 4, "bedrooms": 2, "bathrooms": 1,
    "amenities": ["pool", "kitchen", "air_conditioning", "wifi"],
    "images": ["https://…"],
    "host": { "name": "Marko", "isSuperhost": false },
    "price": { "currency": "USD", "nightlyPrice": 303, "totalPrice": 2122, "nights": 7 }
  },
  "meta": { "platforms": ["booking"], "cached": false, "partial": false, "warnings": [] }
}

4. get_price

A real price quote for one listing, for your dates and occupancy. Pass the platform-native id returned by search_stays (numeric on Airbnb and Vrbo, a slug string on Booking.com and Google Hotels) or a full listing URL. The response is always a real numeric price or a typed error - never a different property's price.

Response (trimmed):

{
  "data": {
    "platform": "booking",
    "listingId": "abramovic2",
    "currency": "EUR",
    "nightlyPrice": 303,
    "totalPrice": 2122,
    "fees": { "cleaning": null, "service": null, "taxes": null },
    "nights": 7,
    "occupancy": { "adults": 2, "children": 2, "childAges": [8, 13] },
    "source": "booking",
    "url": "https://www.booking.com/hotel/hr/abramovic2.html"
  },
  "meta": { "platforms": ["booking"], "cached": false, "partial": false, "warnings": [] }
}

5. compare_prices

The flagship tool. Compare one property's price across booking sites in a single call, resolved through the Google Hotels backbone. The response carries the individual offers plus StayingAPI-computed min and median as first-class fields, so you read the cheapest and typical cross-site price without re-deriving them.

Response (trimmed):

{
  "data": {
    "property": "Hotel X, Sibenik",
    "checkIn": "2026-07-13",
    "checkOut": "2026-07-20",
    "currency": "EUR",
    "min": 2122,
    "median": 2151,
    "offers": [
      { "ota": "booking.com", "totalPrice": 2122, "currency": "EUR", "url": "https://…" },
      { "ota": "expedia", "totalPrice": 2180, "currency": "EUR", "url": "https://…" }
    ]
  },
  "meta": { "platforms": ["google"], "cached": false, "partial": false, "warnings": [] }
}

6. get_reviews

Normalized, paginated reviews for one listing on one platform. Native rating scales are preserved and echoed alongside each rating (Airbnb, Vrbo and TripAdvisor use 5; Booking.com, Expedia and Hotels.com use 10) and never silently rescaled, so a 9 and a 4.5 are not accidentally compared.

Response (trimmed):

{
  "data": [
    {
      "platform": "booking",
      "listingId": "abramovic2",
      "reviewId": "r987",
      "rating": 9, "ratingScale": 10,
      "title": "Perfect family stay",
      "text": "Spotless apartment a short walk from the old town…",
      "author": "Jane D.", "date": "2026-05-10", "tripType": "family",
      "language": "en", "ownerResponse": "Thank you!",
      "liked": "Location and cleanliness", "disliked": null
    }
  ],
  "meta": {
    "platforms": ["booking"], "cached": false, "partial": false,
    "pagination": { "limit": 20, "cursor": null, "nextCursor": "eyJ…", "hasMore": true },
    "warnings": []
  }
}

7. get_job

Poll a long-running scrape that was returned as an async job. When a request is projected to run longer than roughly 8 seconds it returns a job handle; the agent calls get_job until status is completed or failed. Polling always costs 0 credits - the underlying work is billed once, on successful completion.

Response while running, then on completion (trimmed):

// While running - polling is free
{
  "data": { "jobId": "job_3kf…", "status": "running", "pollUrl": "/v1/jobs/job_3kf…", "estimatedSeconds": 12 },
  "meta": { "creditsCharged": 0, "platforms": ["vrbo"] }
}

// On success - the payload arrives in data.result, in the same unified schema
{
  "data": { "jobId": "job_3kf…", "status": "completed", "result": [ /* the endpoint's payload */ ] },
  "meta": {
    "platforms": ["vrbo"], "currency": "USD",
    "platformResults": [ { "platform": "vrbo", "status": "ok", "cached": false, "count": 15 } ],
    "warnings": []
  }
}

A failed job still returns successfully with status: "failed" and the reason nested at data.error (a { type, code, message, retryable } object). Failed work is free.

Data you get

Every call returns the same unified { data, meta } envelope, whether the property is a hotel or a holiday home:

Coverage: Booking.com, Airbnb, Vrbo and Google Hotels - hotels and short-term rentals in the same schema, so one integration covers both halves of the accommodation map.

Use Cases & Prompts

Pricing

This README stays number-free on paid pricing by design, so it never drifts from the live pricing page. MCP and REST draw from the same single credit balance - there is no separate MCP wallet or MCP pricing.

Sign up · View pricing

Troubleshooting

Branch on error.type (the class, mapped 1:1 to the HTTP status) and then on error.code (a stable, finer-grained reason). Every error also carries a requestId, a retryable flag, and a docUrl.

Also available as a REST API

Building an app or a backend instead of an agent? The same data ships as a plain JSON REST service.

Base URL: https://api.stayingapi.com/v1. The seven MCP tools map 1:1 to /v1/search, /v1/availability, /v1/listing/{platform}/{id}, /v1/price, /v1/price-compare, /v1/reviews and /v1/jobs/{jobId}.

Connect

• Website: stayingapi.com
• MCP guide: stayingapi.com/docs/mcp
• Docs: stayingapi.com/docs · OpenAPI: api.stayingapi.com/openapi.json
• Workflows & recipes: stayingapi.com/workflows
• Status: status.stayingapi.com
• Contact: hello@stayingapi.com
• Listed in: awesome-agent-plugins, a directory of verified Agent Plugins on the open standard, maintained by the same team

Related repositories - part of the StayingAPI open resource set: hotel-api · airbnb-api · booking-com-api · vrbo-api · google-hotels-api · travel-api · travel-workflows · travel-skills

MCP Registry

Published to the official Model Context Protocol Registry as:

com.stayingapi/hotel-vacation-rental-mcp

Registry manifests in this repo: server.json (MCP registry) · smithery.yaml (Smithery) · glama.json (Glama).

FAQ

Is there an Airbnb or Booking.com API in 2026? Not one you can sign up for and start using today. Airbnb has no open public API - access runs through its partner programs, which are aimed at approved software partners and property-management systems, not general developers. Booking.com's Demand API is similar: it requires a partner or affiliate agreement and an approval process before you get credentials. Vrbo sits inside Expedia Group's partner API program, and Google Hotels data flows through hotel partner feeds rather than a self-serve developer API. That gap is why "airbnb api" is one of the most-searched, least-served queries in travel tech. StayingAPI is the self-serve alternative: sign up, get a key with 300 free credits and no card, and query accommodation data across all four sources through one REST API or this MCP server, with every response normalized to the same schema.

What is a hotel or accommodation MCP? A hotel or accommodation MCP is a Model Context Protocol server that exposes accommodation data as tools an AI agent can call directly: search stays, check day-by-day availability, fetch full listing detail, quote a real price for specific dates, compare that price across booking sites, and read normalized reviews. Instead of you copying results into a chat, the assistant fetches live data mid-conversation and reasons over it, which means it can chain steps on its own: search, then price the best candidate, then summarize its reviews. This one is hosted at https://mcp.stayingapi.com/mcp, speaks the Streamable HTTP transport, and exposes seven read-only tools covering Booking.com, Airbnb, Vrbo and Google Hotels in a single unified schema. It works with Claude, ChatGPT, Cursor, Claude Code and any other MCP client, and authenticates over OAuth 2.1, so no API key is ever pasted into the agent and nothing here can book, cancel or change a reservation.

What is the difference between the StayingAPI REST API and this MCP server? They are two front doors to the same service, not two products. The REST API is for apps, backends and data pipelines: you send a bearer key, you get JSON, you control the loop. The MCP server is for AI assistants and agents: you add one URL, authorize once over OAuth 2.1, and the seven tools become things the model can call on its own. Underneath they are identical - the same validation, the same adapter and cache pipeline, the same unified schema, the same error taxonomy, and the same single credit balance. There is no separate MCP wallet or MCP pricing. The only real difference is rate limiting: REST is metered per key, MCP per user. Pick REST when you are writing the code, MCP when the model is.

How is this different from a single-platform API or a scraper? A single-platform API gives you one source's view, so cross-platform questions ("is this villa cheaper on Booking.com or Vrbo?") are unanswerable by construction. A SERP scraper hands you a raw page snapshot that you parse and normalize yourself, with no rating scale, no canonical amenity taxonomy, and no guarantee the row you parsed is the property you asked about. This server returns four sources in one schema, keeps native rating scales explicit instead of silently rescaling them, and ships compare_prices, which returns each site's offer plus a computed lowest and median. It also refuses to guess: if a listing's canonical identity does not match what you asked for, you get an identity_mismatch error rather than a plausible-looking wrong property. Failed, empty and blocked calls are never billed.

How do I compare hotel prices across Booking.com, Airbnb and Vrbo with AI? Connect this MCP server to your assistant, then ask in plain language: "compare this hotel's price for 13-20 July across booking sites". The model calls compare_prices with the property name and your dates. It comes back with each site's total in your chosen currency plus StayingAPI-computed min and median fields, so the assistant can tell you not just the cheapest offer but how far below typical it sits. For short-term rentals where the same physical property is listed on several sites under different names, the usual flow is search_stays to find candidates, then get_price per platform for an exact quote at your dates and occupancy. Add the auto-invoke rule from Quick Install and your assistant will reach for these tools on its own whenever you mention a stay.

Can an AI agent search vacation rentals and short-term rentals, not just hotels? Yes, and that is deliberate - it is why this is a hotel and vacation rental MCP. Hotels arrive via Google Hotels and Booking.com; short-term rentals via Airbnb and Vrbo. Crucially they come back in the same object shape: a city-centre hotel room and a rural holiday villa both return property type, bedrooms, bathrooms, max occupancy, a canonical amenities list, a guest rating with its explicit scale, and a price with a fee breakdown. So an agent can answer "what sleeps 4 with a pool near Split, hotel or apartment, whichever is cheaper" in one pass, instead of you running two integrations and reconciling them. search_stays accepts a platforms[] filter and a propertyType[] filter, so you can narrow to just rentals, just hotels, or let both compete.

What data does it return? Every tool returns the same { data, meta } envelope. Property records carry identity (platform, native listing id, canonical URL), core facts (property type, bedrooms, bathrooms, max occupancy, geo), a guest rating with an explicit ratingScale so a 9-out-of-10 is never confused with a 4.5-out-of-5, a canonical amenities list unified across all four platforms, host details, full-resolution photo URLs, and pricing with nightly, total, nights and a fees breakdown. Availability returns per-day available, minNights, checkIn, checkOut and bookable. Reviews return rating on its native scale, title, text, author, date, trip type, language, owner response, and liked/disliked. The meta block always reports the request id, per-platform status and cache flags, pagination cursor, and any warnings, so you can tell a partial result from a complete one.

How do I add it to Claude, ChatGPT or Cursor? Claude Desktop and Claude Web: Settings, Connectors, Add custom connector, then paste https://mcp.stayingapi.com/mcp and complete the OAuth popup. Claude Code: claude mcp add --transport http stayingapi https://mcp.stayingapi.com/mcp, then run /mcp and authorize. ChatGPT: enable Developer Mode under Settings, Apps & Connectors, Advanced settings, then Settings, Connectors, Add MCP server, paste the same URL and complete the OAuth prompt. Cursor: use the one-click badge at the top of this README, or add {"mcpServers":{"stayingapi":{"url":"https://mcp.stayingapi.com/mcp"}}} to ~/.cursor/mcp.json for every project, or to .cursor/mcp.json for just one. Any other MCP client works too - add the same URL as a Streamable HTTP server, and clients that support Dynamic Client Registration will negotiate OAuth without further configuration. Setup takes under a minute in every case, because the server is hosted: there is nothing to install, no runtime to manage, and no package to keep updated. Full per-client detail is in Quick Install, and only configurations verified against the official docs are listed there, since a wrong config block costs you the one install attempt you get.

How much does it cost? Signup gives you 300 free credits with no card, and stay_test_ sandbox keys return deterministic fixtures at zero cost forever, so you can build and test the whole integration before spending anything. After that it is credit-based, with current per-call costs and plan pricing on the pricing page - this README deliberately carries no paid numbers so they can never go stale here. Three things are always free regardless of plan: failed, empty and blocked calls (creditsCharged is 0), get_job polling, and every sandbox call. MCP and REST draw on the same single balance, so adding the MCP server does not create a second bill. Live credit unlocks after you verify your account email; the sandbox works fully before that.

Is StayingAPI affiliated with Airbnb, Booking.com, Vrbo or Google Hotels? No. StayingAPI is an independent service. StayingAPI is not affiliated with, endorsed by, or sponsored by Airbnb, Booking.com, Vrbo, or Google Hotels. These are trademarks of their respective owners, used here descriptively to indicate the data sources StayingAPI's API and MCP server can query. We are not a partner, a reseller, or an authorized integration of any booking platform, we hold no agreement with any of them, and none of them endorse, review or approve this project. Platform names appear throughout this README, in the platforms[] enum, and in tool parameters purely as descriptive labels for which data source a given call targets, in the same way a price-comparison tool names the shops it checks. StayingAPI does not book, cancel or modify reservations: all seven tools are strictly read-only, annotated readOnlyHint: true, and there are no write tools at all. If you need to actually reserve a stay, follow the canonical listing url returned with every property and book on that platform directly.

Get your key — 300 free credits, no card · MCP docs · Pricing · Status

_{Released under the MIT License. StayingAPI is an independent service and is not affiliated with or endorsed by any booking platform. Platform names are trademarks of their respective owners.}