Connect Booking Review to Claude Code in 2026

To connect Booking.com reviews to Claude Code, you add an MCP server or a CLI skill that can read them, because Booking.com's own Claude connector does not support Claude Code and its documentation does not mention reviews. If you run the property, the official route is Booking.com's Guest Review API; if you do not, you are choosing between third-party tools, and Booking.com's terms bar automated access without its written permission.

This guide covers the five routes we found, what each one returns according to its own documentation, the exact Claude Code commands, and where the terms question sits. Every vendor fact below comes from a page we fetched on 5 October 2026. Where a vendor does not publish something, we say so.

What Booking.com offers officially

The Booking.com connector for Claude

Booking.com's connector was added to Claude's connector directory in April 2026. The listing on claude.com carries an "Anthropic verified" badge, names Booking BV as the maker, and shows a single tool, accommodations_search. Lowyat.NET reported on 24 April 2026 that the new batch of connectors was rolling out across all Claude plans, with mobile support in beta.

Booking.com's developer documentation is clear about the limits. The connector searches hotels, apartments and other stays, filters by accommodation type, facilities, star rating and meal plan, and returns summarised results in chat. It works on Claude Web, Claude Desktop, Claude Mobile and Claude Cowork, and Claude Code is listed as not supported. It does not complete bookings, manage reservations, access your Booking.com account or process payments, and the page does not describe returning guest reviews.

So the official connector answers "find me a hotel" in the Claude app. It does not answer "pull the last 500 reviews for this property into my repo."

The Guest Review API, for properties

Booking.com's Guest Review API is the sanctioned way to get review data out programmatically. Per the developer documentation, it retrieves reviews by property ID or review ID, returns overall scores for the property, and lets the property post replies. It uses token-based authentication, and your machine account needs permission to access the review API.

Two restrictions matter. First, access runs through Booking.com's connectivity programme: Elfsight's guide to Booking.com API access states that there is no public Booking.com API and that access is granted only to Connectivity Partners. Second, the documentation says reviews retrieved through the API "may not be used on your public page" and that the API "is for internal use by properties."

That fits an owner or operator analysing their own guest feedback. It does not fit an investor researching a hotel they do not own. If you do have access, connecting it to Claude Code means wrapping the API in a small MCP server of your own, since we could not find a Booking.com MCP server for the review API.

Third-party routes into Claude Code

Three tools document Booking.com review access from Claude Code or another MCP client. None of the pages we fetched states that the vendor holds Booking.com's written permission, so read the terms section below before you pick one.

Printing Press: the booking-com CLI and skill

Printing Press is an open-source library of command-line tools built for AI agents. The Booking.com entry credits Matt Van Horn as creator and Trevin Chow as a contributor. The project README listed 573 CLIs across 22 categories when we checked, and several entries note that they rely on unofficial or sniffed APIs. The library's discovery skill is also listed on skills.sh, a third-party skills directory, and we cover the wider project in our Printing Press Claude skill guide.

The Booking.com entry installs a CLI plus an agent skill with one command:

npx -y @mvanhorn/printing-press-library install booking-com

The binary is booking-com-pp-cli. Its reviews command returns review text, score, traveler type, language, stay date and reviewer country, with filters for score band, language and traveler type. A reviews stats command runs a local SQL group-by over synced reviews and returns counts and median score per bucket. The documented example:

booking-com-pp-cli reviews stats --slug auliviaopera --country fr --by score-band,language,traveler-type --agent

No API key is needed. The documentation says the tool parses Booking.com's server-rendered pages and JSON-LD, uses a Chrome TLS fingerprint to get past the site's AWS web application firewall, and keeps data in a local SQLite database. It lists its own caveats, including that reviews pagination beyond the first page falls back to static HTML. For Claude Desktop it ships an MCPB bundle; in Claude Code you use the skill and the CLI.

Apify: Booking.com MCP Server by axly

Apify hosts a Booking.com MCP server published by a community developer, axly. Apify's own documentation says the billable events on a pay-per-event Actor are defined by the Actor's creator. The listing exposes eight tools, two of them for reviews: get_hotel_reviews, which returns a review score breakdown and featured guest reviews, and get_attraction_reviews.

The listing says it calls "the official Booking.com Android API" with no HTML scraping and no login, authenticating anonymously through the app's device-token flow. Pricing is pay per event at $0.02 per completed tool call. The server runs in Apify's Standby mode, which keeps it running as a web service, and Apify's Standby documentation says requests authenticate with your Apify API token in an Authorization header.

StayAPI: hosted Booking.com MCP

StayAPI publishes a hosted MCP server with five Booking.com tools: booking_search_hotels, booking_lookup_destination, booking_hotel_details, booking_hotel_reviews and booking_hotel_rooms. The reviews tool supports sorting and pagination. It requires an API key, the same key works for StayAPI's REST API, and signup includes 50 free requests with no credit card.

StayAPI's pricing page did not show plans when we fetched it. A FlightPowers comparison updated on 29 September 2026 lists StayAPI at $49 for 1,500 requests, $149 for 10,000 and $450 for 100,000, and describes it as best for enriching listings you already know. FlightPowers discloses that it sells a competing Booking.com data API.

The routes side by side

Read Booking.com's terms before you connect anything

Booking.com's terms of service, effective 25 August 2025, are direct on this point. Section A15.2 says you may not access, copy or scrape anything on the platform using "any robot, spider, scraper, other automated means, or automated assistants," including AI-powered assistants that work through your browser, "without the prior, express written permission of Booking.com." Section A15.3 says Booking.com monitors visits and may restrict automated systems and AI-powered assistants that access the platform without permission.

The clause names AI assistants explicitly, so it reaches an agent reading pages on your behalf, not only bulk scrapers. We are not your lawyers, and this is not legal advice. The practical steps are simple: ask any vendor in writing whether it has Booking.com's permission, keep third-party review data out of anything you publish, and get counsel to sign off before that data goes into an investment committee memo or a lender deck.

Setting it up in Claude Code

Claude Code adds a remote MCP server with one command. Anthropic's documentation gives this form for an HTTP server that needs a header:

claude mcp add --transport http <name> <url> --header "Authorization: Bearer your-token"

StayAPI documents its own version of that command:

claude mcp add --scope user --transport http stayapi https://api.stayapi.com/mcp --header "X-API-Key: YOUR_API_KEY"

For the Apify server, combine the Standby URL from the listing with the token header from Apify's documentation:

claude mcp add --transport http booking https://<your-actor-standby-url>/mcp --header "Authorization: Bearer YOUR_APIFY_TOKEN"

Scope decides who sees the server. Per Anthropic's docs, local is the default and loads in the current project only, project writes a .mcp.json file you can commit for the team, and user loads in all your projects. Run claude mcp list to confirm the server is registered, and use /mcp inside a session to check its status. Claude Code asks for approval before it uses project-scoped servers from a .mcp.json file.

One warning applies to every route here. Anthropic's documentation says to verify you trust each server before connecting it, because servers that fetch external content can expose you to prompt injection. Guest reviews are free text written by strangers. Treat them as data, keep the agent's write permissions narrow while it reads them, and do not give a review-reading session credentials to systems that matter.

What review data is good for in a hotel deal

For an operator, the use is plain: track response times, count recurring complaints, and see whether fixes show up in scores. On the buy side, reviews are a cross-check on the management story, not a valuation input. A seller deck that says a refurbishment lifted guest satisfaction should show up as a shift in the score distribution after the works date. A claimed move toward business travel should show up in the traveler type mix.

The useful cuts are the ones the Printing Press stats command already exposes: score band, language and traveler type, counted over time. Three cautions keep the analysis honest. Small samples move a median a lot, so report the count next to every score. Language and traveler mix shift with season and channel strategy, which can move scores with no change in the product. And a review is one guest's view of one stay, so a pattern in reviews belongs on the question list for management, not in the verdict.

We cover how this fits a wider diligence workflow in AI for financial due diligence, and how deal teams run it across a portfolio in AI for private equity.

Buying a seat versus owning the system

For a one-off look at a single hotel, a third-party MCP server and an afternoon in Claude Code is enough, once the terms question is settled. For a recurring process, such as monthly review monitoring across a portfolio or a standing diligence checklist, the trade-off changes. You want data you are permitted to use, a fixed analysis you can audit, and outputs that land in your own systems.

That is the work consultance.ai does. We build custom finance AI systems, the first one is live in about six weeks, and the client owns the code. If a review feed belongs inside a deal desk system rather than a chat window, book a call and we will tell you plainly whether it is worth building. More free Claude Code setups are in the library.

Sources

• Booking.com Developers: About Booking.com Connector, fetched 5 October 2026
• Claude: Booking.com connector listing, fetched 5 October 2026
• Lowyat.NET: Claude Introduces New Connectors For Spotify, Booking.com, And More, fetched 5 October 2026
• Booking.com Developers: Guest Review API, fetched 5 October 2026
• Elfsight: How to get and use Booking.com API, fetched 5 October 2026
• Booking.com: Terms of Service, fetched 5 October 2026
• Printing Press: Booking.com CLI, skill and MCP server, fetched 5 October 2026
• GitHub: mvanhorn/printing-press-library, fetched 5 October 2026
• skills.sh: printing-press-library skill, fetched 5 October 2026
• Apify: Booking.com MCP Server by axly, fetched 5 October 2026
• Apify Docs: Actor Standby mode, fetched 5 October 2026
• Apify Docs: Actors in Store, fetched 5 October 2026
• StayAPI: Booking.com MCP Server, fetched 5 October 2026
• FlightPowers: The best hotel data APIs in 2026, fetched 5 October 2026
• Anthropic: Connect Claude Code to tools via MCP, fetched 5 October 2026