What is google-search-console-mcp?

An MCP server for Google Search Console, the service that tells you how Google sees your site — which queries brought people to it, which pages are indexed, and what it made of your sitemaps.

It covers three Google APIs, because Search Console alone is not enough to set a property up: Search Console v1 for properties, sitemaps, analytics and URL inspection, Site Verification v1 for proving ownership, and the Indexing API v3 for change notifications.

Why

Search Console's web interface is fine for looking. It is a poor place to do anything repetitive: adding a property means clicking through a verification wizard, submitting forty sitemaps means submitting them forty times, and answering "which of these two hundred pages is not indexed" means inspecting them one at a time.

Those are exactly the jobs an MCP client is good at — and the ones where getting the details wrong is easy and invisible. A property spelled https://example.com instead of https://example.com/ is a different property that answers 403. A seven-day range computed in UTC covers six days of data. A totals row averaged across query CTRs is off by an order of magnitude. This server gets those right so the conversation above it does not have to.

What it will not do

• Make Google crawl something. Nothing can. request_indexing exists because the API does, and it only has an effect on JobPosting and BroadcastEvent pages; every description involving it says so.
• Report live page state. inspect_url returns what Google last indexed, not a fresh fetch. A page fixed an hour ago still shows the old verdict.
• Place a verification token for you. Creating a DNS record or editing a template is a human step by design. setup_site gets you the exact record to paste and picks up again afterwards.
• Show you today's numbers. Search Console finalises data two to three days behind. dataState: "ALL" includes the incomplete recent days; nothing includes the missing ones.

What it is careful about

Four operations cannot be undone and are two-step, requiring a token that only ever appears in a previous tool result: removing a property (which discards roughly sixteen months of history Google will not give back), removing a sitemap, giving up ownership, and replacing the owner list — the last one because the list you pass is the complete list afterwards, so it removes everyone else.

Everything the APIs return is marked as untrusted content. Search queries are strings the public typed into Google; page titles and crawl diagnostics come from whoever runs the crawled site.

And the OAuth scopes follow the tools: narrowing the tool list narrows what the credential is used for. See Security.