yandex-metrika-mcp

Install

npx skillfoxx add mcp/expremiental-yandex-metrika-mcp

claude mcp add --transport http yandex-metrika-mcp https://mcp.getaiaiai.ru/yandex-metrika

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "type": "http",
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

code --add-mcp '{"name":"yandex-metrika-mcp","type":"http","url":"https://mcp.getaiaiai.ru/yandex-metrika"}'

{
  "servers": {
    "yandex-metrika-mcp": {
      "type": "http",
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

codex mcp add yandex-metrika-mcp --url https://mcp.getaiaiai.ru/yandex-metrika

[mcp_servers.yandex-metrika-mcp]
url = "https://mcp.getaiaiai.ru/yandex-metrika"

gemini mcp add -s user -t http yandex-metrika-mcp https://mcp.getaiaiai.ru/yandex-metrika

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "httpUrl": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "serverUrl": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "type": "streamableHttp",
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "mcpServers": {
    "yandex-metrika-mcp": {
      "type": "streamable-http",
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "mcp": {
    "yandex-metrika-mcp": {
      "type": "remote",
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

{
  "context_servers": {
    "yandex-metrika-mcp": {
      "url": "https://mcp.getaiaiai.ru/yandex-metrika"
    }
  }
}

https://mcp.getaiaiai.ru/yandex-metrika

This is third-party code. Review the repository files before installing.

What it does

The server answers questions about Yandex Metrika analytics in plain language: how many visits this week, top sources for the month, today's mobile share. Instead of dozens of narrow tools, it gives just two: list_counters returns the counters available to the token, and query runs an arbitrary Reporting API call with parameters for metrics, dimensions, date range, filters, sort and row limit, covering visits, users, pageviews, bounce rate, sources, devices, geography and UTM tags. The main way to use it is a ready hosted HTTP service at mcp.getaiaiai.ru: no app registration or token retrieval needed, sign-in happens through Yandex OAuth right when the connector is added, and the service is explicitly stated as read-only, writing nothing to the Metrika account. For those who want to keep data under their own control, a Python self-host option with your own OAuth token is available.

Who it is for. For people who want to ask Yandex Metrika analytics questions in plain language without setting up an app or storing tokens.

Example request

How many visits were there this week, and where did most of the traffic come from?

Limitations

The hosted option depends on the third-party aiaiai service and its infrastructure, not just the Yandex API. The query tool requires knowing the exact Reporting API metric and dimension names; there are no ready report presets. The self-host option requires Python 3.10 and uv.

How to disable. Remove the connector from your client's settings (Claude.ai: Settings, Connectors), or remove the server from your MCP client config if self-hosted, and revoke access in your Yandex account settings.

MCP

Transport

http, stdio

Authentication

OAuth

Security check

• Reads site analytics data, including visitor sources and geography
• The hosted option routes login through the third-party aiaiai platform on top of the Yandex account

README in short

The bilingual README presents the server as a way to ask Metrika analytics questions in plain language, with no dashboards, SQL or key management. The main path is the aiaiai hosted service with a one-minute connection for Claude.ai, Claude Code, Claude Desktop and Cursor via Yandex OAuth, with the service explicitly noted as read-only. A table of two tools, list_counters and query, details the latter's parameters: metrics, dimensions, date range, filters, sort, limit, with examples of common metrics and dimensions. A collapsible developer section covers a Python self-host option with uv, your own OAuth token, an HTTP run mode, and embedding into your own backend via an injectable token resolver. MIT license.

Related