TheSEOCentral.

Agent access · Model Context Protocol

Connect an AI agent to The SEO Central

A public, read-only MCP server — add it once; no browser, no per-session setup.

Most websites are built for humans with eyes and a mouse. The Model Context Protocol (MCP) is the open standard that lets AI assistants — Claude, ChatGPT, Cursor and others — call tools on an external service in a structured way. The SEO Central publishes a small MCP server so an agent can query our already-public content directly, instead of scraping pages.

The endpoint

https://www.theseocentral.com/mcp

Read-only · no authentication · JSON-RPC 2.0 over HTTP (Streamable HTTP).

Add it to your AI client — one click

Use a button for Cursor or VS Code, or copy the URL/command for Claude:

Add to Cursor Add to VS Code

Claude (Desktop or web): copy the URL above, then add it as a custom connector in Claude’s Settings → Connectors (authentication: none).

Claude Code:

claude mcp add --transport http seocentral https://www.theseocentral.com/mcp

How it works

This is a server-side MCP server, not a browser plugin. An MCP client adds the URL above once; from then on it can list and call the server’s tools whenever it needs them — no page open, no per-session pairing.

Your AI client (Claude, ChatGPT, Cursor …)
        │  speaks MCP (JSON-RPC 2.0) over HTTPS
        ▼
   https://www.theseocentral.com/mcp   ← runs on our server
        │  executes the read-only tool
        ▼
   returns published data back to your agent

The protocol handshake is standard: the client calls initialize, then tools/list to see what’s available, then tools/call to run one. Our server implements exactly that, and every call is rate-limited and recorded.

What the server exposes

All tools are read-only, over content that is already public on the site:

ToolWhat it does
list_blog_postsRecent published posts (optional limit 1–50).
search_blogKeyword search across the blog (query, optional limit).
get_blog_postOne post by slug, including the full article text.
list_glossary_termsAll glossary terms with short definitions.
get_glossary_termOne definition by term name or slug, with related terms.
list_workshopsPublished workshops — title, price, format, schedule.
search_faqSearch (or list) the technical SEO FAQ (query, optional).
get_site_infoStructured overview of the site and its sections.

Source: the same public content behind /blog, /glossary, /technical-seo-faq and /workshops.

Test it in Claude

Claude Desktop — add the server to your config, then restart Claude:

Config file: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) · %APPDATA%\Claude\claude_desktop_config.json (Windows)

{
  "mcpServers": {
    "theseocentral": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://www.theseocentral.com/mcp"]
    }
  }
}

Claude.ai (web, Pro/Team) — Settings → Connectors → Add custom connector → paste https://www.theseocentral.com/mcp. That’s the whole setup. Then ask Claude: “Using the theseocentral tools, list the latest blog posts.”

Test it in ChatGPT / GPT agents

ChatGPT (Developer mode / Connectors) — in Settings → Connectors, enable developer mode and Add a custom MCP server with URL https://www.theseocentral.com/mcp and authentication set to None. Then, in a chat, turn the connector on and ask it to “list blog posts from The SEO Central.”

OpenAI Agents / Responses API — point a hosted-MCP tool at the same URL:

{
  "type": "mcp",
  "server_label": "theseocentral",
  "server_url": "https://www.theseocentral.com/mcp",
  "require_approval": "never"
}

MCP connector support in ChatGPT is rolling out by plan/region; if you don’t see the option yet, the Claude or Cursor routes work today.

Test it in Cursor / other clients

{ "mcpServers": { "theseocentral": { "url": "https://www.theseocentral.com/mcp" } } }

What the agent gets back

For transparency, here is real data the tools return (shown for humans). Everything below is already public on the site.

list_blog_posts / search_blog / get_blog_post

list_glossary_terms / get_glossary_term

  • 301 Redirect
    A 301 redirect is a permanent redirect that tells browsers and search engines a URL has moved, transferring the bulk of its ranking signals…
  • 404 Not Found
    A 404 Not Found is an HTTP status code returned when a server has no resource matching the requested URL, telling both users and search cra…
  • AI Overview
    An AI Overview is an AI-generated summary that Google displays at the top of search results for certain queries, citing sources drawn from …
  • Answer Engine Optimization (AEO)
    Answer engine optimization is the practice of structuring content so it can be directly quoted as a complete, accurate answer by AI chat as…
  • Canonical Tag
    A canonical tag is an HTML element that tells search engines which URL is the preferred, master version among a group of duplicate or near-…
  • Canonicalization
    Canonicalization is the process search engines and site owners use to choose one authoritative URL when multiple addresses return the same …
  • ccTLD
    A ccTLD is a country-code top-level domain, such as .de for Germany or .jp for Japan, that gives search engines a strong signal a site targ…
  • Click Depth
    Click depth is the number of clicks required to reach a specific page starting from a website's homepage, which strongly affects how often …

list_workshops

search_faq

get_site_info

Returns a structured overview: name, description, founder, and links to the Blog, Glossary, Technical SEO FAQ, Workshops and About sections.

Is this safe?

  • Read-only. The server only returns content that is already public. It cannot write, delete, submit forms, sign in, or reach the admin or database.
  • Public by design. Because it exposes nothing private, it needs no login — that is what removes the setup friction. Anything private or write-capable would be gated behind a key before being exposed.
  • Rate-limited & logged. Each call is capped per IP and recorded, so usage is visible and abuse is contained.

How agents discover it

MCP has no web-wide auto-discovery yet — a client is always told the URL once (the steps above). To make that easier, this server advertises itself two ways:

  • A machine-readable manifest at /.well-known/mcp.json (name, endpoint, transport, and the available tool).
  • A pointer in our llms.txt so LLM-aware crawlers see the endpoint.

Beyond that, public MCP directories (the official MCP registry, and indexes like PulseMCP and mcp.so) are where people browse for servers to add.

How this was built & references

  • Model Context Protocol — the open standard this implements: modelcontextprotocol.io (transport & JSON-RPC methods per the MCP specification).
  • This server is a small custom PHP implementation of MCP’s Streamable-HTTP transport (initialize / tools/list / tools/call), serving one read-only tool over The SEO Central’s published content. Endpoint: https://www.theseocentral.com/mcp.