MCP
MCP Server
PageMore page actions
The YGO Console MCP server lets an AI assistant read your YGO projects and hotels and send hotel feedback to the YGO team on your behalf. You connect the server to Claude, ChatGPT or a coding assistant, sign in with your YGO account, and ask in plain language: which projects you have, which hotels a project holds, which of them are inactive, what YGO answered to your feedback. The assistant sees only what your account can see in the console.
Server URL: https://mcp.ygo.ai/mcp
Who Can Use It
- Any YGO Console user can sign in. Project, hotel and feedback access needs the Editor role or higher in the organization that owns the project.
- Sign-in is OAuth with your YGO account. There is no API key to copy and no password typed into the assistant.
- Hotel and feedback requests name a project, and YGO checks your role in its organization on every call. A list of your feedback can also leave the project out: it then covers every project you can reach.
The Tools
An assistant that is connected to the server receives a small set of tools. You never call a tool by name: you ask a question, and the assistant picks the tool and fills in the parameters. The connector offers seven tools. Five are on every connection; send_hotel_feedback and reply_to_feedback need write access, described under Hotel Feedback below.
| Tool | What it does | Writes anything? |
|---|---|---|
list_projects |
The projects your account can reach, with id, name, organization and status. | No |
list_hotels |
One page of a project's hotels: id, name, country, star rating, status, text status and GIATA id. Filter by status, by tag name or by a name search (a number matches the GIATA id). Hotels you deactivated in the project are left out unless you ask for them. Paged, up to 100 hotels per page. | No |
list_my_feedback |
The feedback you sent, in every project you can reach or in one, newest activity first: the project, the hotel, your text, its status (open, acknowledged, resolved, rejected or superseded), the latest YGO answer and the number of replies. Filter by hotel, by status, by activity since a date or by whether a thread has replies, or open one thread to read its replies, oldest first. 10 threads per call by default, up to 50, taken from your 100 newest comments across those projects; filter by project, hotel or date to reach older ones. | No |
list_hotel_requests |
The hotels your team asked YGO to add to one project, with each request's state (requested, approved, rejected or added), the hotel's name, city, country and URL, and the reason when YGO declined one. Filter by state. 10 requests per call by default, up to 50. Available once YGO has enabled hotel requests for the project. | No |
send_hotel_feedback |
Sends a comment about one hotel in a project to the YGO team. A call not marked as confirmed is a preview and writes nothing. A call marked as confirmed stores the comment. The assistant is told to preview first, but YGO cannot enforce it; see Hotel Feedback below. | A new comment; needs write access |
reply_to_feedback |
Adds a reply in one of your own feedback threads: to your comment or to a reply in it (from YGO staff, a colleague or you). It previews and confirms the same way as send_hotel_feedback. |
A new reply; needs write access |
request_capability |
Tells the YGO product team about something you asked YGO for that no YGO tool can do, for example prices, availability, bookings, exports or editing hotel data. Sends a short summary of the request and your account email to YGO and changes nothing in your projects, hotels or feedback. The assistant uses it once, in the turn in which it tells you YGO cannot do it, and hands you a reference such as REQ-.... Quote the reference when you follow up with integrations@ygo.ai. The summary is recorded and shared with the YGO team under your name, so do not put secrets or guest data in what you ask. |
A request to YGO only |
Questions that work today:
- "List my YGO projects."
- "Show the hotels in Sunshine Travel."
- "Which hotels in Sunshine Travel are inactive?"
- "Find hotels in Sunshine Travel with 'beach' in the name."
- "Show hotels tagged 'family' in Sunshine Travel."
- "How many hotels does Sunshine Travel have?"
- "What feedback did I send in Sunshine Travel, and has YGO answered?"
- "Tell YGO that the description of Seaside Resort in Sunshine Travel still mentions the closed pool." (write access)
- "Reply to YGO's answer on my Seaside Resort feedback: thanks, that fixes it." (write access)
Hotel Feedback
send_hotel_feedback, reply_to_feedback and list_my_feedback carry a comment about a hotel to the YGO team and bring the answer back. A comment does not edit the hotel: the YGO team reads it and can answer in the same thread.
- Write access. Sending and replying need write access, which the sign-in screen asks you to grant or decline. If you decline, the other five tools keep working, the two write tools are not offered, and the assistant can only tell you to reconnect and grant write access. To grant it later, disconnect and reconnect the connector.
- Preview, then confirm. A send or reply call posts nothing unless the assistant marks it as confirmed. Without that mark it returns where the text would land and the exact text that would be stored: for a comment the organization, project and hotel, for a reply the message being answered and its hotel. The assistant is instructed to confirm only after your explicit yes, and YGO cannot check that you said it, so keep tool approval on and read the text in the confirming call before you approve it.
- One call, one message. Each confirmed call stores a new comment or reply. YGO does not detect a repeat, and the assistant cannot edit or withdraw what was sent, so check the preview instead of sending again.
- Plain text. The text is stored as plain text: whitespace is trimmed, line breaks and tabs become spaces, a vertical bar becomes a slash, a backtick becomes an apostrophe, invisible characters are removed, and long token-like strings are replaced with
[credential removed]. That filter is a safety net, not a complete one: it does not catch passwords, short keys or personal data, so do not put secrets or guest data in feedback. - Who reads it. A comment or reply you send is stored in your project under your account and reaches the YGO team with your email address. The YGO team can read and answer it, and so can people who work with your project's feedback in the YGO console or through the API. Everyone who reads it there, a project API key with feedback read access included, sees your email as the author.
- Your threads only. Through the assistant you read the comments you sent, not those of your colleagues, and you reply only in your own threads. Inside your thread you see the replies that have not been withdrawn, each marked with whether you wrote it and whether it came from a customer account or YGO staff, or as unknown when YGO cannot tell. No author names, emails or ids are returned. A comment YGO has marked superseded stays in the list, but its replies cannot be opened and it takes no new reply.
What It Does Not Do
- It does not change hotel content, prices, availability, settings or users.
- It does not show hotels or projects outside your organization.
- It does not show a hotel's descriptions or enriched attributes.
- It does not show comments your colleagues sent, and it does not edit or withdraw a comment or reply.
Limits
| Limit | Value |
|---|---|
| Tool calls | 60 per minute per user |
| Capability requests | 10 per hour per user |
| Confirmed feedback sends and replies | 10 per minute per user, the two tools combined |
| Feedback text | 2,000 characters per comment or reply |
| Replies in one thread | 200 |
A call over the limit answers with an error the assistant reports back to you; nothing is queued.
Security
- OAuth, not credentials. Sign-in runs through YGO's identity provider: you sign in on YGO's own sign-in page, never inside the assistant, and the assistant receives an access credential that expires within minutes. Its client also keeps a longer-lived refresh credential so you are not asked to sign in on every call.
- Your access, nothing more. You see the projects of the organizations where you are an Editor or higher. If your role in an organization changes or is removed, the assistant loses that access on its next request.
- Audit trail. Successful calls, including the preview of a feedback post or reply, are recorded on a best-effort basis with the user, the tool, the client and, where the call names one, the project. A list of your feedback across projects is recorded once for each project it read, and not at all when you can reach no project. YGO staff can see that record.
- Names and texts are data. Hotel and project names are stripped of hidden characters and labeled untrusted in the results that carry them. The feedback texts in your threads carry the same label. The assistant is therefore told to treat all of them as data, not as instructions. This is a notice to the assistant, not a guarantee of its behavior: keep tool approval on in your assistant and check what it proposes before you confirm a write.
- Disconnecting. Disconnect the connector in your assistant's settings. The assistant stops using your sign-in and asks you to sign in again next time. The access credential is short-lived and expires on its own. Removing a user's role in the organization stops their project, hotel and feedback access at once. To end the sign-in itself, contact
integrations@ygo.ai.
Need help? Reach our integration team at integrations@ygo.ai.
