Google Search Console MCP: The Server Google Never Shipped

50+ official servers, none for Search Console

Google has an official MCP server for Gmail. For Calendar. For Maps, Drive, BigQuery, Cloud Run and Firestore. Google Analytics has an official one. Search Console, the one dataset every SEO opens first, has nothing.

That absence is the answer to the search that probably brought you here. There is no official Google Search Console MCP server, there is no announced plan for one, and every result promising you one is somebody else's code wrapped around the public API. Verified August 2026.

If MCP itself is the unfamiliar word in that paragraph, our glossary explains what the Model Context Protocol is in two minutes. The short version: it is the open standard that lets Claude, ChatGPT and other agents call tools over a URL, and it is why this page is about servers rather than browser plugins.

What people mean by a Google SEO MCP

"Google SEO MCP" is really three different searches, and they end in three different places.

• An MCP server made by Google for SEO. Does not exist. The closest official servers are Google Analytics, which has no search query data, and BigQuery, which only helps if you already export your data there.
• An MCP server for your Google search data. Exists in two forms: community wrappers you host yourself, and vendor servers with Search Console integration built in. That choice is most of this page.
• Any SEO MCP server that covers Google results. A different question with a longer answer. We ranked all nine vendor servers in our SEO MCP server comparison.

Route 1: self-host a community wrapper

The open-source route works, and for some readers it is the right call. Two projects dominate:

• ahonn/mcp-server-gsc, the most widely packaged wrapper, serving search analytics straight from the API
• AminForou/mcp-gsc, which adds regex filtering, larger row limits and a quick-wins pass

Setup is the same shape for both: create a Google Cloud project, enable the Search Console API, generate credentials, grant them access to your property, and configure your MCP client to launch the server. An evening's work if you are comfortable in a terminal, a lost weekend if you are not.

What self-hosting actually costs

The software is free. The ownership is not, and it is the part the README never prices in.

• You are the host. The server runs on your machine or your infrastructure, and it is down when they are.
• You hold the credentials. A Google service-account key or OAuth secret sits wherever you put it, and rotating it is your job.
• Nobody owes you a fix. When Google changes the API, an unpaid maintainer decides when, and whether, the wrapper catches up.
• The agent gets rows, not answers. A wrapper passes through what the API returns. Joining it to analytics or crawl data is a second project.

Route 2: a vendor server with GSC built in

The second route is a hosted server from a vendor whose product already integrates Search Console. You skip the Google Cloud console entirely: the integration is built in, the vendor keeps it running, and the connection is a URL plus a browser sign-in.

This is the gap CrawlRaven's MCP server was built into. Eight read-only tools serve your Search Console data joined with GA4 and a 200-point crawl: totals and trends, queries and pages with position and CTR, ranked opportunities with evidence, and your target-keyword workspace. The endpoint is https://mcp.crawlraven.com/mcp, authentication is OAuth authorization code with S256 PKCE, and no token is ever pasted into a config file.

The join is the practical difference between the two routes, not the hosting. A wrapper can tell an agent that clicks fell. A joined server can put the ranking drop, the engagement dip and the crawl finding on the same page in one tool call, which is what turns "what happened?" into "here is what to fix".

The honest catches: MCP access requires a paid plan, from $49 at launch as a one-time licence, so free accounts cannot connect. And every tool is read-only by design, so an agent cannot change anything in your account, which you may count as a feature.

One caution about the wider market: most SEO MCP servers do not touch this problem at all. Ahrefs, Semrush, Serpstat and DataForSEO serve their own indexes. Excellent for competitor research, but none of them can read what your site actually ranks for, because none of them can see your Search Console account.

Route 3: skip MCP and use the filters

The least glamorous route is underrated. If you have one question a week and it fits a filter, Search Console's own UI answers it with no server, no OAuth and nothing to maintain.

The query regex filter does most of the work an agent would. Our free GSC regex generator builds the pattern for you: brand vs non-brand splits, question queries, long-tail filters, all pasteable straight into the performance report.

The moment this route stops working is the moment your questions become conversations: "compare this month to the update", "now just the money pages", "which of those also have a crawl issue?". Filters answer questions. They do not follow up.

The three routes, compared

Same decision, condensed. The full comparison against every vendor server lives in the MCP server roundup; this table is just the Search Console question.

What to ask once you are connected

Whichever server you connect, the setup steps are the same two: add the URL, approve in the browser. The per-client walkthroughs, including ChatGPT's developer-mode gate and the four failures that explain most bad connections, are in our client-by-client connection guide.

The better question is what to ask once the tools appear. Prompts that earn their keep on day one:

Which pages lost the most clicks in the last 28 days compared to the previous 28, and what position changes explain it?

List queries where I rank between position 5 and 15 with real impressions. Which pages should each one point at, and what would you change first?

Of the pages losing clicks, which ones also have crawl or engagement issues, and what is the single highest-impact fix this week?

That last prompt only works on a joined server. Against a bare wrapper the agent will answer from ranking data alone, which is a guess wearing a confident tone. The complete tool catalog, scopes and limits are in the MCP server documentation.

Which route I would pick

For a personal site and a comfortable terminal, self-host ahonn/mcp-server-gsc and enjoy it. The setup is a fair evening's project, the price is right, and your only user is you.

For client work, buy the supported server. Not because the wrapper fails, but because "the maintainer has not updated it yet" is not a sentence you want to say on a client call, and because the answers get materially better when search, analytics and crawl data arrive joined.

About the Author

Aditi Chaturvedi

15+ years of growing SaaS websites through SEO | Author, 200-Point Audit Checklist

Aditi has spent 15+ years helping SaaS companies scale organic traffic through technical SEO and content strategy. She is the author of the CrawlRaven 200-Point Audit checklist used by agencies and in-house teams to systematically improve search performance.