Skip to content
Docs / Guides

Connect an AI agent with the Pagereckon MCP server

The Pagereckon MCP server lets Claude, Cursor or any other MCP client read your site's Pagereckon data. Your agent can then answer questions such as which pages get the most real visitors, which AI crawlers read your posts, and which pages to fine tune next.

It runs on your own computer with one npx command and talks only to your own WordPress site. Nothing in it edits your content.

The Pagereckon team Updated
Mcp at a glance, from Pagereckon

What the MCP server can do

The server, pagereckon-mcp, offers six tools. Two of them work with the free plugin. The other four need Pagereckon Pro, because they use Search Console data and fine tuning marks.

Tools in the Pagereckon MCP server
ToolNeedsWhat it returns
`top_pages`FreePages with the most visits from verified humans, with their sources
`ai_bots`FreeAI crawlers and assistants that read the site, and the pages they read
`fine_tune_candidates`ProPages with many impressions and too few clicks, missed clicks, and search terms no page covers
`page_queries`ProOne page's Google and Bing search terms, its fine tuning mark and its last report
`mark_tuned`ProRecords a change, so the page gets a before and after report on its control day
`control_due`ProPages whose control day has come, with the report

Only mark_tuned writes anything, and it writes a Pagereckon mark, not a change to a post. So an agent can read and suggest, while you stay in charge of every edit.

Create an application password for MCP

The server signs in to the WordPress REST API with an application password. WordPress has offered these since version 5.6, and by default only on sites served over HTTPS.

  1. In WordPress, open Users, Profile for an administrator account.
  2. Under Application Passwords, enter the name pagereckon-mcp and add it.
  3. Copy the password WordPress shows. It only appears once.

The user must be an administrator, because Pagereckon's reports are admin only. You can revoke the password from the same screen at any time, and the agent then loses access.

Add the MCP server to your client

Most MCP clients, such as Claude Desktop and Cursor, read a JSON config. Add a pagereckon entry with your site URL, the user name and the application password:

MCP client config (JSON)
{
  "mcpServers": {
    "pagereckon": {
      "command": "npx",
      "args": ["-y", "pagereckon-mcp"],
      "env": {
        "PAGERECKON_URL": "https://example.com",
        "PAGERECKON_USER": "admin",
        "PAGERECKON_APP_PASSWORD": "xxxx xxxx xxxx xxxx xxxx xxxx"
      }
    }
  }
}

Claude Code in one line

In Claude Code, one command adds the same server:

Claude Code
claude mcp add pagereckon -e PAGERECKON_URL=https://example.com -e PAGERECKON_USER=admin -e PAGERECKON_APP_PASSWORD="xxxx xxxx xxxx xxxx xxxx xxxx" -- npx -y pagereckon-mcp

The server needs Node.js 18 or newer. After you restart the client, ask it something simple, for example which AI crawlers read your site last week.

Security and HTTPS for the MCP connection

The application password travels only from your computer to your own site. It never goes to pagereckon.com or anywhere else. Still, always use an https:// URL, because a plain HTTP connection would expose the password on the network.

In addition, keep the config file private, since it holds the password. If you share a project folder, put the values in environment variables instead. Finally, use a separate application password for each tool, so you can revoke one without breaking the others.

A typical MCP session

First, the agent calls control_due to report on pages whose control day has come. Next, it calls fine_tune_candidates and explains the top pages: a CTR problem for pages in the top 10, a position problem for positions 11 to 30. Then it reads one page with page_queries and writes a proposal for you.

You decide what to change and publish it yourself. Only after the change is live does the agent call mark_tuned, so the before and after report can measure it. The fine tuning skill packs this workflow into a file your agent can follow.

MCP questions

Does the MCP server work with the free plugin?

Yes, partly. top_pages and ai_bots work with the free plugin. The other four tools need Pro, because they use Search Console data.

Can the agent edit my posts through MCP?

No. Nothing in the server edits content. The only write is mark_tuned, which records a fine tuning mark in Pagereckon.

Which MCP clients can use it?

Any client that runs local MCP servers over stdio, such as Claude Desktop, Claude Code and Cursor.

Sources and references

Free plugin, optional Pro

See your own AI traffic in ten minutes

Install the free plugin, point it at your access log and read today's humans, bots and AI visits. Then add Pro when you want Search Console and the before and after reports.