TourScanner

TourScanner AI guide

How to use the TourScanner API: search tours and activities, check live availability, and book them. It also covers the MCP server, if you want an AI agent to do the same thing.

It's a REST API (JSON over HTTPS, OpenAPI 3.1) for tours, activities and attractions worldwide. The catalog is about a million activities from 14 providers, behind one set of endpoints. This repo is the practical side of the docs: what to call, in what order, and the mistakes that cost people an afternoon.

Reference docs live at tourscanner.ai/docs, and the machine-readable spec is at tourscanner.ai/openapi.json.

Get a key

Keys are issued by hand. Email api@tourscanner.ai with what you're building and which cities matter to you. Send it in the X-API-Key header on every request. Without one you get a 401.

export TOURSCANNER_API_KEY=your_key

First request

curl -s "https://tourscanner.ai/api/search/activities?city=rome&q=colosseum&limit=3" \
  -H "X-API-Key: $TOURSCANNER_API_KEY"

You get back activities, plus total, page and some facets. Each activity has an id, a name, a rating, and a price of the form { "currency": "EUR", "price": 5400 }.

Prices are integers in minor units, so 5400 is 54.00. This applies to the min_price and max_price filters too.

The booking flow

Five calls, in this order:

1. GET /api/search/activities to find something.
2. POST /api/availability/slots for the real slots and prices on a date.
3. POST /api/booking/options for the ticket variants and booking questions.
4. POST /api/booking/hold reserves the slot and returns a Stripe checkoutUrl. The card is authorized, not charged.
5. POST /api/booking/confirm after the customer pays. This confirms with the supplier and captures the payment.

Details and request bodies are in docs/booking.md. A runnable version of steps 1 to 4 is in examples/node/book.mjs.

Examples

Node, Python, Ruby, Go and curl. Each does the same three things: search, check availability, and run the booking flow up to the Stripe link. None needs a package beyond the language's standard library (Node 18 or newer).

Run the Go ones from examples/go. The curl scripts are in examples/curl.

They read your key from TOURSCANNER_API_KEY. Set TOURSCANNER_BASE_URL to point them at a different host. Run book without OPTION_CODE first: it prints the options and stops, so you can pick a code. On the production host a full book run places a real hold and creates a real Stripe checkout.

Missing your language? The API is plain JSON over HTTPS, so any HTTP client works. The OpenAPI spec above will generate a client for most languages.

For AI agents

There's a remote MCP server at https://tourscanner.ai/api/mcp. It exposes 12 tools covering search, availability, checkout links, and booking status. Setup for Claude, Cursor, VS Code and others is in docs/mcp.md.

Things that trip people up

• Cents, not euros. Every price field is in minor units.
• productOptionCode comes from booking/options. The hold fails without it, and the codes differ per activity.
• adults is required in passengers, and it has to be at least 1.
• Phone numbers need the country code, like +442079460000. Anything else is rejected.
• Stripe sessions expire after 30 minutes. After that the hold is released.
• Check availability right before booking. Search results come from our catalog, and slots come live from the supplier. They can disagree.
• Not every provider books in-house. For the others you hand the traveller off to the partner instead. Adding providers=2,3,7 to a search limits it to the ones we book directly. The MCP get_activity tool reports this per activity as bookableByUs.
• Rate limits return 429 with a Retry-After header. Honor it.

Environments

The production host places real holds and real Stripe checkouts. If you want a sandbox on provider test rails, ask for one when you get your key.

License

MIT for the code in this repo. The catalog content you get from the API is covered by the terms.