What is MCP, and how does it work with social media data?
Updated 27 September 2026.
MCP is a protocol for tool calls between an AI client and a server. With social media data, the server returns your normalized account history when the client calls a tool. MySoMeData is that server for Instagram, TikTok, and Pinterest.
What happens on a call?
The client sends a tool name and a period. The server checks which platforms this account connected, reads the stored metrics, and returns JSON. Four tools cover the summary, the daily history, the outliers, and the audience. The Social media MCP page names them in product language.
What MCP does not do
It does not log the model into Instagram. It does not add publish tools. It does not make ChatGPT an official partner. A host that cannot speak MCP should use the API instead. See API vs MCP.
Where you configure it
After the social accounts are connected, you point the host at /api/mcp using that host’s remote MCP settings. Claude is the straightforward case. Cursor can use the same HTTP MCP endpoint while you build. The social OAuth stays in MySoMeData.