Server

The marketing platform for your AI agents.

An MCP (Model Context Protocol) server for the PageRankCafe API. It gives an AI agent tools to read and post link ads, banner ads, YouTube video ads, and press releases/blog articles on PageRankCafe, check credit and referral balances, read per-placement performance, and (on paid tiers) pull platform-wide ad performance insights.

Getting a key

1. Create a PageRankCafe account at https://pagerankcafe.com.
2. Upgrade to a paid membership (Bronze/Silver/Gold) at https://pagerankcafe.com/membership-info if you want to use the write and insights tools. Read tools work on a free account too.
3. Generate an API key at https://pagerankcafe.com/apiKeys.

Account creation and key generation are both manual, human steps on the website — there is no programmatic signup.

Configuration

The server reads its configuration from environment variables:

Install

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "pagerankcafe": {
      "command": "npx",
      "args": ["-y", "pagerankcafe-mcp"],
      "env": {
        "PAGERANKCAFE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Claude Code

claude mcp add pagerankcafe --env PAGERANKCAFE_API_KEY=your-api-key-here -- npx -y pagerankcafe-mcp

Cursor

Add to your Cursor MCP settings (~/.cursor/mcp.json or the project's .cursor/mcp.json):

{
  "mcpServers": {
    "pagerankcafe": {
      "command": "npx",
      "args": ["-y", "pagerankcafe-mcp"],
      "env": {
        "PAGERANKCAFE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Tools

Every write tool and the insights tool require a paid membership (Bronze/Silver/Gold); the API returns an error for free accounts. Read tools work on any valid key, including free accounts.

Full request/response detail for each underlying endpoint is in the OpenAPI spec at https://pagerankcafe.com/api/spec (rendered at https://pagerankcafe.com/api/docs).

Error handling

Every tool returns the API's JSON response as the tool result on success. On a non-2xx API response, the tool returns an MCP tool error (isError: true) carrying the API's error message. There are no retries and no silent fallbacks — a failed call always surfaces as a failed tool call.

Development

npm install
npm run build   # compiles src/ to dist/ with tsc
npm test        # builds, then runs the stdio smoke test (no network required)

The test starts the built server as a subprocess over stdio (using StdioClientTransport from @modelcontextprotocol/sdk), lists its tools, and checks the full tool list is present — no real API key or network access is used.