# Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard that lets an assistant such as Claude, Cursor or ChatGPT call tools and read data from a server. For SEO work it means your Search Console data, your analytics or a report can become tools the agent calls while it edits your pages. Before connecting a server, check its scope, where its token lives and who runs it.

Updated 2026-09-23 · Source: https://porteur.ai/glossary/model-context-protocol

## How it works

Anthropic introduced MCP in November 2024 as an open standard. It defines one way for an assistant, the client, to ask a program, the server, what it offers and then use it. A server describes its tools, such as "query search analytics" or "get page", with their inputs. The assistant decides when to call one, the server runs it and returns the result, and the assistant carries on with its task.

Servers run in one of two ways. A local server runs on your machine and talks to the assistant over stdio, its standard input and output. A remote server runs elsewhere and is reached over streamable HTTP, usually with a key or a sign-in. The same server works with any client that speaks the protocol, which is why one server can serve Claude Code and Cursor alike.

## What it means for SEO work

A coding agent sees your code and nothing about how search engines treat it. MCP is how the missing half gets in. With a Search Console server connected, the agent can look up which queries /pricing is shown for, then rewrite that page's title in the same session. With a report exposed over MCP, it can read one finding and make the change it describes.

| Connected server | What the agent can then do |
| --- | --- |
| Search Console | Find pages at positions 4 to 20, compare periods, see which queries a page is shown for |
| Product analytics | See which pages bring visitors who sign up, not only visitors |
| A report with findings | Read a finding whole, with the page and the change, and apply it |

The server does not make the data better. Search Console through MCP still omits rare queries, still lags a few days and still knows nothing about who else ranks. What changes is that the agent no longer works from guesses.

## Adding a server

In Claude Code, servers are added with claude mcp add. Cursor reads them from .cursor/mcp.json.

```bash
# a local server (stdio)
claude mcp add <name> -- <command> [args]

# a remote server (streamable HTTP)
claude mcp add --transport http <name> <url> --header "Authorization: Bearer <token>"
```

Servers added with project scope in Claude Code are stored in .mcp.json at the repository root, a file you commit. That is useful for sharing a server with a team, and the wrong place for any secret.

## What to check before connecting a server

- Scope: does it ask for read-only access? A server that reads your search data has no need to write to it.
- Tokens: where does it store them? Keep them in environment variables or a file outside the repository, never in .mcp.json.
- Operator: who runs it? A remote server holds your token on someone else's machine. A local one runs code you should have read.
- Tools: what can it do? List its tools after connecting, and remove any server you do not use.

## Questions

### What is an MCP server?

A program that exposes tools and data to an assistant using the Model Context Protocol. It can run on your machine or remotely. The assistant calls its tools when a task needs them.

### Is MCP only for Claude?

No. Anthropic introduced it as an open standard, and other assistants and editors, Cursor among them, act as clients. A server written once works with any of them.

### Is MCP useful for SEO?

Yes, when the agent that edits your site needs figures it cannot see in the code. Connecting Search Console or a report lets it choose and change pages from real data instead of general advice.

### Is it safe to connect an MCP server?

As safe as the server and its permissions. Grant read-only scopes, keep tokens out of committed files, and connect only servers whose code or operator you trust.

## Read next

- [Google Search Console MCP: giving Claude or Cursor your search data](https://porteur.ai/guides/google-search-console-mcp): What a Google Search Console MCP server is, the three ways to get one, setup for Claude Code and Cursor, the prompts worth asking, and how to keep it safe.
- [Claude Code for SEO: what it can fix in your repository, and what it cannot see](https://porteur.ai/guides/claude-code-for-seo): Use Claude Code for SEO: what it fixes in your repo, the search data it lacks, how to bring that data in, a worked session and rules for CLAUDE.md.
- [A Claude SEO skill for your repository: what it should hold, and one to copy](https://porteur.ai/guides/claude-seo-skill): A complete Claude Code SEO skill to paste into .claude/skills: what it checks in your repo, the findings it writes, what it cannot see, and how to adapt it.
- [The Search Console API: pull your own search data](https://porteur.ai/guides/search-console-api): What the Google Search Console API gives you, how to authorise it, the request you send, quotas, BigQuery export, and three automations to ship.
- [Agentic SEO](https://porteur.ai/glossary/agentic-seo): Agentic SEO is search work done by AI agents that read data, edit pages and publish in a loop. What it means for a small site, and its traps.
- [SEO AI agents: what an agent can do for a small site, and where it goes wrong](https://porteur.ai/guides/seo-ai-agent): What an SEO AI agent is, the loop that works for a small site, where agents go wrong, and a table of what to automate, review and never delegate.

Before connecting anything, see what your site's searches look like: the free check reads your site, the searches around it and the rivals on them from a URL in about thirty seconds. Free check: https://porteur.ai/