How do I give Claude or ChatGPT access to cruise data?

What does the cruise data MCP server do?

The Cruise Itinerary MCP server exposes the Cruise Itinerary API as read-only tools that Claude, ChatGPT and other MCP clients can call: cruise lines, ships, ports, sailings, itineraries and weekly fares from 43 lines. It is served at https://cruise-itinerary.com/mcp over Streamable HTTP, and each call counts toward the plan of the signed-in account.

The MCP authorization specification has clients discover how to sign in from the server's protected resource metadata, so a client needs only the URL. The server lists six read-only tools, each gated by the same plan as the REST endpoint it wraps:

• resolve_entity: turns a port, ship or line name, misspelt or old, into the id the other tools take. Every plan.
• cruise_price_index: the weekly Cruise Price Index; the latest 12 weeks on every plan, full series by line, cabin and segment on Business.
• search_sailings and get_itinerary: upcoming sailings with their latest weekly fares, and day-by-day itineraries. Developer and above.
• get_price_history and port_month_capacity: every weekly fare snapshot of a sailing, and monthly ship-days and berths at a port. Business and above.

A tool above your plan returns a message with the upgrade link instead of data.

Which cruise lines can the assistant look up?

Through the MCP server an assistant can look up all 43 cruise lines in the Cruise Itinerary register, 61,010 future sailings from 2026-09-29 to 2030-10-02. MSC Cruises, Viking River Cruises and Hurtigruten have the most sailings. TUI and Marella, AIDA, Saga, Fred. Olsen and Hapag-Lloyd are not covered.

The full list, with ship counts and how many sailings carry a fare, is on the coverage page.

How do I add cruise data to Claude?

In Claude on the web or the desktop app, go to Customize, then Connectors, choose Add, then Add custom connector, and enter a name and https://cruise-itinerary.com/mcp. Claude detects the sign-in method; choose to sign in now. Custom connectors are available on Free, Pro, Max, Team and Enterprise plans; Free allows one.

1. Open Customize > Connectors and select + Add, then Add custom connector.
2. Name it "Cruise Itinerary" and paste https://cruise-itinerary.com/mcp as the server URL.
3. Keep the detected authentication, choose Sign in now, and sign in with your Cruise Itinerary email.
4. To use an API key instead, add it under Request headers as Authorization: Bearer ci_live_….

On Team and Enterprise plans an owner adds the connector under Organization settings, and members connect it from Customize. Anthropic's custom connector guide has the current screens.

How do I connect Claude Code to cruise data?

Claude Code adds a remote MCP server with one command: claude mcp add --transport http cruise-itinerary https://cruise-itinerary.com/mcp. Run /mcp inside Claude Code to sign in through the browser, or pass an API key with --header "Authorization: Bearer" followed by the key, which suits scripts and CI.

claude mcp add --transport http cruise-itinerary https://cruise-itinerary.com/mcp

Then type /mcp in Claude Code and follow the browser sign-in. With a key instead:

claude mcp add --transport http cruise-itinerary https://cruise-itinerary.com/mcp \
  --header "Authorization: Bearer $CRUISE_ITINERARY_KEY"

The commands are from the Claude Code MCP documentation. Claude Desktop's claude_desktop_config.json is for local servers; add remote ones through Connectors.

How do I add cruise data to ChatGPT?

ChatGPT connects to remote MCP servers through developer mode, available on the web for Plus, Pro, Business, Enterprise and Education accounts. Turn it on under Settings, Security and login, then open ChatGPT Plugins, select the plus button and create a developer-mode app with the server URL, signing in through OAuth.

1. In ChatGPT on the web, open Settings > Security and login and turn on Developer mode. A workspace admin may need to allow it.
2. Go to Plugins, select + and create a developer-mode app.
3. Enter a name, a short description and https://cruise-itinerary.com/mcp, then complete the OAuth sign-in.

OpenAI's developer mode guide documents OAuth and no-authentication servers; it does not describe custom headers, so use OAuth sign-in in ChatGPT.

How do I call the MCP server from an API?

Both the OpenAI Responses API and the Anthropic Messages API can call a remote MCP server for you. OpenAI takes an mcp tool with server_url and an authorization token. Anthropic takes an mcp_servers entry with url and authorization_token, plus an mcp_toolset tool and the mcp-client-2025-11-20 beta header.

OpenAI Responses API tool (OpenAI MCP guide):

{
    "type": "mcp",
    "server_label": "cruise_itinerary",
    "server_url": "https://cruise-itinerary.com/mcp",
    "authorization": "ci_live_...",
    "require_approval": "never"
}

Anthropic Messages API, with the header anthropic-beta: mcp-client-2025-11-20 (MCP connector docs):

{
    "mcp_servers": [
        {
            "type": "url",
            "url": "https://cruise-itinerary.com/mcp",
            "name": "cruise-itinerary",
            "authorization_token": "ci_live_..."
        }
    ],
    "tools": [
        {
            "type": "mcp_toolset",
            "mcp_server_name": "cruise-itinerary"
        }
    ]
}

Use a key from your account. Keys start with ci_live_ and are shown once.

What does a cruise data tool call return?

A tool call returns JSON with a data block and a meta block carrying the load date. Asked to resolve "Icon of the Seas" on 2026-10-03, resolve_entity returned one exact match, the ship icon-of-the-seas with confidence 1, from the register loaded on 2026-09-28. The assistant then passes that id to other tools.

The real response, as the server returned it on 2026-10-03 (the structuredContent of the result):

{
    "data": [
        {
            "type": "ship",
            "id": "icon-of-the-seas",
            "name": "Icon of the Seas",
            "confidence": 1,
            "match": "exact"
        }
    ],
    "meta": {
        "as_of": "2026-09-28",
        "page": 1,
        "per_page": 1,
        "total": 1,
        "has_more": false
    }
}

A confidence below about 0.6 is a guess, and the server tells the assistant to confirm it with you. Every response carries as_of, so the assistant can say which Monday load an answer comes from.

What should an assistant know about this cruise data?

Assistants should treat fares as weekly Monday snapshots of lead-in fares, not live bookable prices, and read a missing fare as not offered or sold out. Days between embarkation and disembarkation are exact only where the line's published itinerary was fetched; elsewhere they are estimated, and responses say which.

• Capacity is in lower berths, not passengers; ships often sail above or below that figure.
• Day-level port counts are estimates; monthly totals are reliable.
• TUI and Marella, AIDA, Saga and Fred. Olsen are not in the register, and Hapag-Lloyd is listed with no sailings, so UK and German sailings are under-represented.
• Fare history starts on 2026-09-06, so it cannot yet show a full booking season.

The same rules apply to the REST API, and its responses carry the fields an assistant needs to say how sure it is.

Key takeaways

• The Cruise Itinerary MCP server lives at https://cruise-itinerary.com/mcp and accepts OAuth sign-in or an API key as a bearer token.
• In Claude, add it under Customize, Connectors, Add custom connector; Free plans allow one custom connector.
• In ChatGPT, turn on developer mode and create a developer-mode app with the server URL and OAuth.
• Claude Code connects with claude mcp add --transport http, then /mcp to sign in.
• Assistants must treat fares as weekly snapshots and estimated days as estimates, never as live or exact.