Metricool ships an official MCP server, so an AI client can pull your social analytics, find the best posting times and schedule content — all from chat. Because it is a hosted remote server, setup is mostly authentication. Here is how.
Option A — remote URL (recommended)
Metricool runs a hosted MCP endpoint at https://ai.metricool.com/mcp. In a client that supports remote/OAuth MCP, add it as a connector and authorise through the browser — nothing to install. This is the cleanest path and works with Claude (Desktop, Code), Cursor, ChatGPT and others.
Option B — local config with a token
For clients configured via JSON, use your Metricool credentials:
{
"mcpServers": {
"metricool": {
"command": "npx",
"args": ["-y", "mcp-metricool"],
"env": {
"METRICOOL_USER_TOKEN": "your_token",
"METRICOOL_USER_ID": "your_user_id"
}
}
}
}
Find your token and user ID in Metricool account settings. Restart the client; see MCP config file location.
Requirements
The MCP server itself is free, but API access requires a paid Metricool plan. Connect your social accounts in Metricool first — the MCP reads whatever your account can see.
Verify
Ask: "What were my best-performing Instagram posts last month, and when should I post next?" Real metrics + a time recommendation confirm it works.
Common problems
unauthorized— wrong token/user ID, or your plan lacks API access.- No data for a network — that social account is not connected in Metricool.
- No tools in client — for Option B, Node 18+ is required; see Cursor MCP not working.
Going further
Metricool pairs with a web-search server for competitive research and a filesystem server for exporting reports. Browse curated loadouts or the content category. Keep credentials safe — see MCP security best practices.