AI Connector · Model Context Protocol
Gymdooh speaks MCP, the standard AI assistants use to reach real systems. Connect it once and Claude, ChatGPT or the assistant your team already uses can read your studios, campaigns and delivery figures — and open a support ticket while you are still typing.
No dashboard, no export, no filter to set. The assistant asks the same API the portal asks, with your permissions.
“Which screens in Munich were offline yesterday?”
Network
“How did the spring campaign deliver last week, by day?”
Delivery
“Why is this campaign not playing yet?”
Readiness
“How much budget is left, and what did it buy?”
Budget
“Open a ticket for the dark screen at the entrance.”
Support
“Pause the ad that is over-delivering.”
Control
The Model Context Protocol is an open standard for letting an AI assistant use an outside system. The assistant is handed a list of things it may do — read your campaigns, open a ticket — and calls them when your question needs them, instead of guessing from what it was trained on.
A connector is one address you add to your assistant once. From then on it is part of the conversation, on your laptop and on your phone alike, and every answer comes from the live platform rather than from a screenshot you pasted in.
In plain language, in the assistant you already use.
Through the same public API the portal uses, signed in as you.
From live data, with the numbers it actually read.
Five minutes, once. You need a Gymdooh account.
In the portal under My Account → API Credentials. You get a client id and a client secret; the secret is shown once. Creating one asks for your password and a code from your authenticator, because the credential outlives the browser session that made it.
Open API CredentialsPick your assistant. The address is the same for all of them:
Works in Claude on the web, in the desktop app and on mobile — the connector is stored with your account, not on one machine.
Custom MCP connectors are offered on some ChatGPT plans only, and OpenAI moves that line from time to time. If you cannot find the setting, your plan does not have it yet — nothing on our side is missing.
Claude Code, Cursor, VS Code and most other MCP clients read a configuration file. Add the server as a remote endpoint:
The client opens a browser window for the sign-in the first time, the same as the two above. Where a client asks for the transport, it is streamable HTTP.
Try “list my Gymdooh accounts” to see that it works, then ask the questions you actually have. The assistant will tell you when it needs a right you do not hold — the platform decides that, not the assistant.
19 operations, each one a call to a documented endpoint of our public API. Nothing is invented on the way.
Each of these is marked as a change, so an assistant that asks before acting will ask.
The connector is not a second way into the platform. It signs in as you and gets exactly your permissions — a studio you cannot see, it cannot see.
Want an assistant with less than your own access? Create a person in your account with fewer rights and make the credential there. A credential always has exactly its owner's permissions.