Skip to content

/ GSC MCP Server

A Google Search Console MCP server, done right.

Google does not publish an official MCP server for Search Console. That SEO Agent is one: a hosted MCP server that gives Claude, ChatGPT, or Cursor direct, read-only access to your real Search Console data through a dedicated set of tools. Connect once with your Google account. Your AI queries your live GSC data itself, no CSV exports, no local server, no Python.

Works in:ClaudeChatGPTCursorAny MCP client
All
the dedicated Search Console tools your AI can call: performance, indexing, sitemaps, and detection
That SEO Agent MCP server
Read-only
Google OAuth scope (webmasters.readonly). The server cannot submit URLs, change properties, or write anything
Google OAuth, read-only scopes
25,000
rows per query through the Search Console API, against the 1,000-row cap in the GSC interface
Google Search Console API, query limits
3+
MCP clients out of the box: Claude, ChatGPT, and Cursor, plus any app that speaks MCP
Streamable HTTP transport, MCP standard

01

Is there an official Google Search Console MCP server?

No. Google does not publish an MCP server for Search Console. What exists is two separate things: the Search Console API, a REST interface you script against, and MCP (the Model Context Protocol), the open standard Anthropic released in November 2024 for letting AI assistants call live tools instead of relying on what you paste into the chat. A Search Console MCP server is the bridge between them: a server that speaks MCP to your AI and calls the Search Console API on your behalf.

That SEO Agent is that bridge, built around Search Console from the start rather than bolted on. It exposes the whole of Search Console as read-only tools, alongside Google Analytics 4, PageSpeed, and a live crawler. When you ask your AI about your rankings, it calls the right tool and answers from your current data, not from a training-cutoff guess.

You could build your own. The Search Console API is open and the MCP SDKs are public. But a production server means handling Google OAuth and token refresh, scoping it to read-only, respecting rate limits, paginating past the interface's 1,000-row cap, hosting the thing, and writing each tool by hand. That SEO Agent is the hosted version of all of that: one connection, nothing to run locally.

The access is strictly read-only. The server authenticates with your Google account using read-only OAuth scopes (webmasters.readonly for Search Console), so it can read your performance and indexing data but cannot submit URLs for indexing, change a property, or alter anything in your account.

2024
the year Anthropic open-sourced MCP. It is now supported natively by Claude, ChatGPT, and Cursor
anthropic.com, Model Context Protocol
MCP is an open standard, not tied to one AI vendor. A server that speaks it works across every compatible client.

Connect your Search Console to your AI in three steps

See pricing

02

The Search Console tools your AI can call

Each tool maps to a real Search Console capability, exposed so your AI can call it directly. Grouped by what they do.

Performance & queries
gsc_search_analyticsgsc_search_analytics_by_indexing_stategsc_search_appearancegsc_page_query_mapgsc_discover_performancegsc_branded_splitgsc_device_gapgsc_country_opportunity
Indexing & coverage
gsc_inspect_urlgsc_bulk_url_inspectiongsc_index_coverage_analysisgsc_sites_health_checkgsc_crawl_freshnessgsc_page_changesgsc_list_properties
Sitemaps
gsc_list_sitemapsgsc_get_sitemapgsc_sitemap_url_inspection
Detection & opportunities
gsc_detect_anomaliesgsc_detect_lost_queriesgsc_detect_quick_winsgsc_detect_cannibalizationgsc_detect_trendsgsc_detect_featured_snippetsgsc_serp_features_gapgsc_rich_results

These are part of That SEO Agent's full toolset, which also covers Google Analytics 4, PageSpeed, technical SEO, and site management.

1
One prompt. Live Search Console data.
From Claude, ChatGPT, or Cursor, ask: "Pull the top 50 queries for yourdomain.com over the last 28 days, flag the ones on page two, and tell me which lost clicks versus last month." The agent calls gsc_search_analytics, gsc_detect_quick_wins, and gsc_detect_lost_queries against your real data and answers, no export, no pasting, no 1,000-row ceiling.
That SEO Agent, Search Console tools

03

How the connection works

That SEO Agent is a remote MCP server over Streamable HTTP. There is no local process to run and no Python environment to maintain. Your AI client connects to a single URL, authenticates once, and the connection persists across sessions, so you are not re-authenticating every time you open a chat.

Authentication has two layers. With your Google account, you grant read-only OAuth access to Search Console and GA4, that is what lets the server read your data. With your MCP client, you authorize the connection once via OAuth 2.0 in the browser, or paste an API key (sea_xxxx format) from the dashboard. Either way, your Google credentials stay with Google.

Because the transport is the standard MCP Streamable HTTP, the same connection works in Claude, ChatGPT, and Cursor, and in any other application that implements the MCP client spec. There is no separate build per client: one server, one URL, every compatible AI.

Once connected, your AI reads each tool's description from the MCP layer and decides which to call on its own. You ask a question in plain language; it picks the right Search Console tool, fetches your live data, and reasons over it.

1
connection, persistent across sessions. No per-conversation setup, no re-auth, no local server
That SEO Agent, remote MCP server
Streamable HTTP is the standard MCP transport, so the same URL works across Claude, ChatGPT, and Cursor.

04

Connect Search Console in three steps

STEP 01
Connect your Google account
Sign in with Google to grant read-only access to Search Console and Google Analytics 4. Your site registers automatically the first time your agent audits it. The agent uses your existing Google OAuth permissions, and your credentials stay with Google.
STEP 02
Authorize your AI client
Connect from Claude, ChatGPT, or Cursor via OAuth 2.0 in the browser, or generate an API key (sea_xxxx format) in the dashboard and paste it into your client's MCP configuration. One connection, persistent access.
STEP 03
Ask about your data
Ask: "What are my top queries this month, and which pages are slipping?" Your AI calls the right Search Console tools and answers from your live data. No exports, no copy-pasting, no row limits.
Sources and references
Official documentation
Model Context Protocol — official documentationmodelcontextprotocol.io
Introducing the Model Context Protocol — Anthropicanthropic.com
Search Console API — Google Developersdevelopers.google.com
Standards and APIs
Search Console API usage limits — Google Developersdevelopers.google.com
Google API OAuth 2.0 scopes (read-only)developers.google.com
MCP transports — Streamable HTTP specificationmodelcontextprotocol.io
/ FAQ

Common questions about a GSC MCP server

Does Google offer an official Search Console MCP server?

No. Google publishes the Search Console API (a REST interface) but not an MCP server. To connect Search Console to an AI assistant over MCP, you either build a server that wraps the API yourself, or use a hosted one like That SEO Agent, which exposes read-only Search Console tools your AI can call directly.

Is the connection read-only? Can it change anything in my Search Console?

It is strictly read-only. The server authenticates with read-only OAuth scopes (webmasters.readonly for Search Console, analytics.readonly for GA4). It can read your performance, indexing, and sitemap data, but it cannot submit URLs for indexing, modify a property, or write anything to your account.

Which AI clients can I use it with?

Any client that supports MCP. That includes Claude (Desktop and Code), ChatGPT, and Cursor, plus any other application implementing the MCP client spec. That SEO Agent uses the standard Streamable HTTP transport, so a single URL-based configuration works across all of them.

Do I need to run a local server or write code?

No. That SEO Agent is a hosted remote MCP server. There is no local process, no Python environment, and no script to maintain. You connect your AI client to one URL, authorize once, and the connection persists across sessions.

How is this different from just using the Search Console API?

The Search Console API is a REST interface you have to script against, parse, and feed into your AI yourself. An MCP server exposes those capabilities as tools your AI calls on its own, in plain language, as part of answering your question. You skip the scripting, the export step, and the work of explaining the data to the model.

Does it get past the 1,000-row limit in the Search Console interface?

Yes. The Search Console interface caps exports at 1,000 rows, but the API returns up to 25,000 rows per query and supports pagination beyond that. Because the agent queries the API directly, your AI works with the full result set instead of a truncated UI export.

MCP server for SEOAnalyze GSC with AIDiagnose a traffic dropMCP setup guide →All research →

/ GSC MCP Server

Connect Search Console to your AI.

Read-only tools. One connection. Your live data.

See pricing

Sign in with Google and query your real Search Console data from Claude, ChatGPT, or Cursor.