SEOforGPT MCP Server

Remote MCP server for SEOforGPT, built with Node.js, TypeScript, @modelcontextprotocol/sdk, and Cloudflare Workers using the streamable-http transport.

Features

• list_projects: list all SEOforGPT projects for the configured API key
• check_brand_visibility: run custom AI visibility checks for a brand across 3 to 5 prompts
• run_project_analysis: run saved prompts for an existing SEOforGPT project
• get_visibility_report: fetch a full previously generated report

Quick start

SEOFORGPT_API_KEY=sgpt_xxx

Install dependencies and build:

npm install
npm run build

Run locally with Cloudflare Workers:

npm run dev

Deploy when you have your Cloudflare credentials available:

npm run deploy

Configuration

The server expects one environment variable:

• SEOFORGPT_API_KEY: your SEOforGPT API key, sent to the upstream API as Authorization: Bearer <SEOFORGPT_API_KEY>

For local or deployed Worker environments, you can set it in wrangler.toml, via wrangler secret put SEOFORGPT_API_KEY, or through your deployment environment.

Endpoints

• GET /: basic server metadata
• GET /health: health check
• POST /mcp: MCP streamable HTTP endpoint

Claude Desktop config

{
  "mcpServers": {
    "seoforgpt": {
      "command": "npx",
      "args": ["-y", "seoforgpt-mcp"],
      "env": { "SEOFORGPT_API_KEY": "sgpt_your_key_here" }
    }
  }
}

Smithery URL

https://seoforgpt-mcp.workers.dev

Tool usage examples

1. List projects

{
  "name": "list_projects",
  "arguments": {}
}

2. Check brand visibility

{
  "name": "check_brand_visibility",
  "arguments": {
    "brand": "Acme",
    "queries": [
      "Best CRM for small businesses",
      "Top alternatives to HubSpot for startups",
      "Which sales tools are best for B2B teams"
    ]
  }
}

3. Run a saved project analysis

{
  "name": "run_project_analysis",
  "arguments": {
    "projectId": "proj_123"
  }
}

4. Get a report

{
  "name": "get_visibility_report",
  "arguments": {
    "reportId": "report_123"
  }
}

Development notes

• The implementation is stateless and creates a fresh MCP server per request, which fits Cloudflare Workers well.
• All tool failures are normalized to { "error": "...", "details": "..." } so the Worker does not crash on upstream API failures.
• The server uses process.env.SEOFORGPT_API_KEY with an env binding fallback for Worker compatibility.