SEOLinkMap MCP Server Complete Guide

Your SEOLinkMap account comes with a private MCP server. Connect it to the assistant you already use, Claude or ChatGPT, and the assistant answers from your own crawl, rankings, traffic and research instead of from training data. It is included on every account.

What Private MCP Access Is

The numbers are measured by SEOLinkMap. The assistant only reads them. Ask which pages lost clicks, what the audit found on a URL, or what the pages outranking you have in common, and the answer is built from the figures on your own screens, with the tool name and the numbers in it. A question the data cannot answer comes back as "no data", not a guess.

Setting Up Your Connection

The MCP server connection uses secure OAuth authorization similar to "sign in with Google" but for your SEOLinkMap account. You'll receive a unique server URL that connects your AI client directly to your private data.

It works with Claude on the web, Claude Desktop, Claude Code, ChatGPT and any other MCP client that supports remote servers. Your server URL is on the dashboard.

The server sees the projects your account can see. In the chat, ask it to list your projects and switch to the one you want; everything after that reads from that project until you switch again.

For detailed step-by-step instructions on connecting your AI tool to MCP servers, see our guide on how to connect your AI chat tool to MCP servers.

Fix Site Issues With Claude Code

Connect the private MCP to Claude Code and one chat can find a problem, edit the code, and confirm the fix. The audit tools say what is wrong, Claude Code changes the files in your repository, and analyze_url_content runs the same audit rules against the live page to tell you whether the change worked.

How To Do SEO With Claude Code is the full walkthrough of this loop, from the first audit call to targeting what the SERP actually rewards.

1. Ask what is broken. "What issues does my site have?" calls get_audit_summary and returns each issue category with a count. "Show me the pages with short titles" calls get_audit_issue_pages and returns the URL list for that one issue.
2. Let Claude Code make the edits. Hand it the page list and the rule those pages failed. It has your templates and content files open, so it edits the source instead of describing what to change.
3. Deploy the change. The recheck fetches the URL over the network, so the edit has to be live before it means anything.
4. Recheck the page. "Re-analyze https://yoursite.com/that-page" runs the audit rules against the page as it serves right now. An empty issue list means it passes every rule that can be checked without a Lighthouse run. Whatever is still listed is what is left to fix, and step 2 starts again.

The recheck measures the page from that fetch, including the redirect chain the URL currently serves. Duplicate title, description and URL findings are the exception, since a duplicate is a property of the whole site and has to compare against the stored crawl.

Nothing from a recheck is saved. get_audit_summary keeps reporting the last scheduled audit until you call queue_crawl, so work through the page list one URL at a time and re-crawl once at the end.

The URL has to be on the active project host exactly. A different subdomain is a different site, so switch_project first if the page lives on one.

What Data You Can Ask About

The fastest question is what do I do next, which reads everything below at once and answers with one action. Everything else here is for when you want the detail behind it.

The private MCP reads your own crawl, your Google accounts, the research and prospecting you have bought, your AI visibility tracking, and the SEO action log. It does not produce numbers of its own, so a question it has no data for comes back as "no data" rather than a guess.

The site crawl covers every page it found, with the title, headings, word count, technical checks, and which pages link to which. It also calculates PageValue, so you can ask which of your pages hold the most link value. Our LinkMap analysis tools show that same structure as a colored map if you would rather look at it than ask about it.

Google Search Console supplies ranking positions and click counts, and Google Analytics supplies traffic. Ninety days is the ceiling on both, and each needs connecting once on the settings page.

Competitor research is the paid part, priced per keyword. It studies the pages currently ranking for a keyword you pick and measures what they have in common.

What The Server Can Do

• Projects: list them, switch between them, create one, and check whether a crawl is running.
• Audit: the issue counts, the pages behind any one issue, and what to do next.
• Pages and links: any page's measurements, what links to it and what it links to, ranked by PageValue.
• Google data: Search Console rankings and clicks, Analytics traffic, both to ninety days.
• SEO action log: read it and add to it.
• Competitor research: your orders, each one's action plan, the correlations behind it, and the keyword and backlink data it holds.
• Live checks: run the audit rules against any page on your site as it serves right now, or against pasted text.
• Orders: queue a crawl, a layout, a research order, a prospector order or AI visibility tracking. Each returns an approval code and waits for a person to approve it before anything is charged.
• Backlink Prospector: your orders, their prospects, and the outreach list to read and update.
• AI visibility: your tracked prompts, their results day by day, and advice on a prompt before you buy it.
• Documentation: searchable from the chat.

Ask In Plain Language

You type an ordinary question and the assistant picks the tool. There is no query syntax to learn and no menu to hunt through.

"What is wrong with this page?" runs the audit rules against the live URL and lists what failed. "Which of my pages have the most internal links pointing at them?" returns your own site ranked by PageValue.

"How did my rankings move last month?" pulls Search Console positions and clicks for that window. Answers come back as sentences with the numbers in them, and follow-ups keep the context, so "why?" and "show me the worst ten" both work.

What The Numbers Mean

Competitor research findings carry two numbers, and both have plain meanings. Correlation runs from -1 to 1 and says how tightly a measurement moves with position: the further from zero, the more the ranking pages agree on it, and the sign tells you which direction the winners lean. Near 0 means no relationship worth acting on.

The p-value says how likely that pattern is a coincidence. Below 0.05 is the usual bar for taking a finding seriously, which means there is less than a 1 in 20 chance of seeing it by luck.

You do not have to read either number yourself. Ask what one means and you get it in sentences, and the findings arrive already sorted into four groups.

Four Ways To Spend Effort

Every measured factor lands in one of four groups, and this is the part most people act on.

CRITICAL is table stakes. Nearly every ranking page has it, so its absence is what holds you back rather than its presence being an edge.

OPPORTUNITY correlates well with ranking and is unlikely to be chance, which makes it the best use of your next hour.

AVOID covers factors with weak correlation and high odds of coincidence. This is where popular advice gets caught not working for your particular keyword.

GAMBLE means the math points somewhere but the evidence is thin. Test it if you have the room, skip it if you do not.

Why Each Keyword Differs

Those four groups are calculated per keyword, because what ranks for one search rarely matches what ranks for another. Word count that matters for a buying query can mean nothing for a how-to query.

That is why an answer here can contradict general SEO advice, and why it can contradict itself across two of your own keywords. What gets measured is the set of pages actually ranking for that one search, not the web in general.

Getting Started with MCP

The platform features page covers what else SEOLinkMap measures, and pricing lists what each plan includes.

For developers interested in implementing their own MCP servers, our MCP Protocol Server Implementation Overview covers how to build one.

If you're considering creating your own MCP service, learn about transforming your website into an MCP service to make your content conversationally accessible to AI assistants.

Start your free SEOLinkMap account today to experience statistical SEO intelligence through conversational AI.