Skip to content

Ask your data: the MCP command center

Natural-language questions over live GSC data, and how it differs from the MCP server.

  • Lesson 2 of 7
  • Intermediate
  • 5 min read
  • Checked against the product 2026-11-03

In shortThe MCP command center (Analytics → Search Console → MCP command center) is a chat box over your live Search Console data. Type a question such as "Which pages lost clicks last week?" and an AI model builds one Search Console query, fetches the rows and answers in plain English with tables. Each message costs credits. Despite the name, it is not the AI Rank Lab MCP server you connect to Claude or Cursor.

What it is, and what it isn't

The command center is for questions that the overview doesn't answer directly: "Which queries containing *pricing* gained impressions this month?", "Show clicks by device for my blog pages", "What are my top 20 queries in India?". You ask in your own words; it turns the question into a Search Console query, runs it on the site you selected and summarises the result.

The name causes confusion. MCP command center is a chat page inside the dashboard. The MCP server is a separate developer feature that lets Claude, Cursor and other AI clients call AI Rank Lab tools from outside the dashboard. They share three letters and nothing else. The MCP server is covered in Connect the MCP server to Claude or Cursor.

Before you start

  • Connect Search Console and pick a site first (previous lesson). Without a selected site the page shows No site selected.
  • Have credits available. Every message you send is charged, whatever the answer turns out to be.

Ask your first question

  1. Open the command centerGo to Analytics → Search Console → MCP command center. Check that the site selector in the top bar shows the property you want: each chat belongs to one site.
  2. Type a specific questionUse the box under Start a Conversation (the placeholder reads *e.g. Why did my blog traffic drop last week?*). Name the metric, the dimension and the period: clicks or impressions, by query, page, country, device or date, over which dates. A new chat session is created automatically.
  3. Read the answer and the numbers behind itWhile it works you'll see Querying Search Console data.... The answer comes back as text with bold key numbers and tables. If the question can't be answered from Search Console data, it says so.
  4. Follow up, rename or deleteAsk a follow-up in the same chat; the last five messages are sent along as context. Use the three-dot menu on a session in Chat Sessions to Rename or Delete it. New Chat starts a clean thread.

How it answers, so you can ask better

Each message runs one Search Console query. The model chooses the dates, one or more dimensions from query, page, country, device and date, and a row limit, then summarises what comes back. If you don't give dates it uses the last 30 days. That has practical consequences:

Works wellWorks badly
"Top 20 pages by clicks in the last 7 days""Why did traffic drop?" with no period: it can describe data, not diagnose causes
"Clicks by device for URLs containing /blog/ last month"Comparing two periods in one message: ask for each period, or use the URL ranking analyzer
"Queries with over 1,000 impressions and CTR under 1%"Anything outside Search Console: GA4 sessions, conversions, AI Overview citations, backlinks
Check one number

AI summaries can misread a table. Before you act on an answer, check one figure against the GSC overview or Search Console itself. If they disagree, ask the question again with explicit dates.

What it costs

The price per message is shown under the session list and in the chat header. The label says *coins*, but it comes out of your normal credit balance.

You pay per message, even for a non-answer

Credits are taken before the AI model runs. A question it can't answer ("that isn't Search Console data"), a Google error or a vague reply still costs the full message price, and there is no automatic refund. Write specific questions, and use the free overview for simple totals.

If you already work in Claude Desktop, Claude Code or Cursor, the MCP server's gsc-performance tool reads the same Search Console data free of credits, and your AI client does the reasoning. The command center is the easier option when you want to stay in the dashboard and keep chat history per site.

Check yourself

1. What is the MCP command center?
2. You ask "How many conversions did my blog drive?" What happens?
3. Which question is most likely to get a precise answer?

0 of 3 answered.

Progress is saved in this browser.

Put the lesson to work

Paste your site and get a free SEO, AEO and GEO audit with the fixes, then come back to the next lesson.