Model Context Protocol · Google Search Console

Google Search Console MCP Server for Claude

A free, hosted Search Console MCP (GSC MCP) that lets Claude query your real search data: queries, pages, clicks, impressions, CTR and average position. Connect once with Google sign-in. No Google Cloud project, service account or local install required.

Status: Available now. Read-only access to Search Console, plus Google Analytics 4 from the same connector.

Connect Search Console to Claude in 1 minute

Open claude.ai → Settings → Connectors.
Click the + next to "Connectors" → Add custom connector.
Paste the server URL below → Connect → sign in with the Google account that has access to your Search Console property and approve read-only access.

You only paste this one URL. Claude discovers the OAuth endpoints automatically. Team or Enterprise workspaces use the same flow under Organization settings → Connectors once an admin enables custom connectors.

Search Console tools Claude gets

Tool	What it does
`list_search_console_sites`	Lists every Search Console property your Google account can access, both domain (`sc-domain:`) and URL-prefix properties.
`query_search_console`	Flexible Search Analytics query by query, page, country, device, search appearance or date, for web, image, video, news or Discover results, up to 1,000 rows per call.
`find_ctr_opportunities`	Query/page pairs with high impressions but lower CTR than expected for their position. Prioritised list for title and meta rewrites.
`find_ranking_opportunities`	"Striking distance" pages ranking in positions 4-15 that could reach the top 3 with on-page work, sorted by impressions.
`get_seo_snapshot`	28-day overview combining Search Console (top queries, pages, countries, devices) with GA4 traffic and channels.

The connector also includes ready-made prompts (CTR audit, rank booster, content gaps, monthly review) and GA4 tools. See the GA4 + Search Console MCP guide for the Analytics side.

What you can ask once it's connected

"Why did my impressions drop last month? Compare pages and queries against the previous period and find when it started."
"Which queries rank on page 2 for pages that could realistically move to page 1?"
"Find queries where two of my pages compete for the same keyword."
"Split my clicks into branded and non-branded queries and show the trend."
"Which countries and devices bring impressions but almost no clicks?"

Copy/paste prompts

Diagnose a traffic or impressions drop

Finds the date, the pages and the queries behind a decline

Using my Google Search Console site, diagnose why my search performance dropped.

1) Pull daily clicks, impressions and average position for the last 6 months and identify when the decline started.
2) Compare the 8 weeks before the drop with the most recent 4 weeks by page and by query (normalise for the different period lengths).
3) List the pages and queries responsible for most of the loss, with their position before and after.
4) Classify the cause for each: ranking loss, lost queries, CTR change, or demand change.
5) Give me a prioritised fix list.

If you need me to pick a Search Console site, list the options first.

Keyword cannibalization check

Queries where several of your pages split impressions

Using query_search_console with dimensions ["query", "page"] for the last 90 days, find queries where two or more of my pages each get at least 10% of the impressions.

For each case show the query, the competing pages, their impressions, clicks and average position, and recommend which page should own the query (merge, redirect, re-target or differentiate).

Striking-distance keywords

Positions 4-15 with on-page fixes

Run find_ranking_opportunities on my Search Console site. For the top 10 pages, list their main queries with position and CTR, then give specific on-page changes (title, H1, intro, missing sections, internal links, FAQ or schema) to move each into the top 3.

Hosted vs self-hosted Search Console MCP

Several open-source Search Console MCP servers exist on GitHub. They're a good fit if you want to run everything locally, but they usually need a Google Cloud project, the Search Console API enabled, and a service account or OAuth client you manage yourself.

	SellOnLLM (hosted)	Typical self-hosted server
Setup	Paste one URL, sign in with Google	Google Cloud project, API keys or service account, local config
Works in Claude.ai (web)	✓	Usually desktop or CLI clients only
Access	Read-only	Depends on the credentials you grant
GA4 in the same connector	✓	Varies
Data stays on your machine	—	✓

Frequently asked questions

What is a Google Search Console MCP?

An MCP (Model Context Protocol) server that exposes Search Console data as tools an AI assistant can call. Instead of exporting CSVs, you ask Claude a question and it queries your clicks, impressions, CTR and positions directly.

Is the Search Console MCP free?

Yes. The hosted connector is free to use. You need a Claude plan that supports custom connectors.

Can it change anything in my Search Console?

No. It requests read-only Search Console and Analytics scopes, so it can't submit sitemaps, remove URLs or change settings. You can revoke access anytime from your Google account's third-party permissions.

Does it support domain properties?

Yes. Both domain properties (sc-domain:example.com) and URL-prefix properties (https://example.com/) work, as long as your Google account has access.

How much history can it query?

Whatever the Search Console API provides: up to 16 months of data, typically delayed by 2-3 days.

Does it work with clients other than Claude?

It's a standard remote MCP server with OAuth discovery at https://www.sellonllm.com/api/mcp. Claude.ai is the supported, tested client; other MCP clients that support remote servers with OAuth may also connect.