Headless SEO Workflows with the Model Context Protocol

A headless SEO workflow is a model that can run a check without a person clicking through a form. The Model Context Protocol is the piece that lists tools, validates arguments, and returns the result as text the model can read. Seoffix ships that server as @seoffix/mcp.

Two ways to connect

Desktop apps that spawn a process use stdio. The process is npx -y @seoffix/mcp once the package is on the npm registry under the Seoffix organization. Until that publish, the server that is already running is the public endpoint https://seoffix.com/api/mcp. Health lives at /api/mcp/health. Legacy SSE is /api/mcp/sse. Streamable HTTP is a POST to /api/mcp. The configuration block for desktop apps is on the API docs.

The five tools

seoffix_check_dns_propagation compares one record type across Cloudflare, Google, and Quad9.
seoffix_whois_lookup returns RDAP registration, expiry, status, and nameservers. It does not say a name can be purchased.
seoffix_pdf_convert extracts text or compresses a PDF on the server. It does not convert to Word.
seoffix_ai_humanizer rewrites prose with the HumanBypass model and does not guarantee a detector score.
seoffix_generate_logo returns an image URL from VectorCraft. It does not search trademarks.

Keys and quotas

Anonymous calls use the hourly limit. A key from Pro or Apex raises that to 1,000 or 2,500 requests a day. The key looks like sfx_live_ plus 48 hex characters. Put it in SEOFFIX_API_KEY for the stdio process. Do not put the key in a prompt, a screenshot, or a shared document.

Smithery and Glama manifests are in the package so those directories can list the server. The protocol surface those clients call is the same five tools, against the same gateway.