GSC Connect - Google Search Console
Connect OpenClaw or another MCP-compatible agent to Google Search Console and retrieve official read-only properties, performance, sitemap, and URL inspection data. Use for clicks,…
2winter
@2winter-dev
What This Skill Does
MCP-compatible service that connects AI agents to Google Search Console for read-only access to property data, performance metrics, sitemaps, and URL inspection results. Uses OAuth to authenticate through Google's official APIs.
Replaces manually exporting Search Console reports or building custom API integrations by giving agents direct access to clicks, impressions, CTR, rankings, and indexing data.
When to Use It
- Check search performance trends for a specific website property over a date range
- Identify top-performing queries and pages by clicks or impressions
- Inspect whether a specific URL is indexed and identify any coverage issues
- List all sitemaps submitted to Search Console and verify their status
- Compare CTR and average position across different countries or devices
- Retrieve daily performance data for automated SEO reporting
Install
$ openclaw skills install @2winter-dev/gsc-connectGSC Connect
Connect the agent to Google Search Console through YUSIHK's public, read-only MCP service. Retrieve remote data only through Google's official Search Console APIs.
Requirements
- A Google account that already has access to at least one Search Console property.
- An MCP client that supports Streamable HTTP and OAuth.
- MCP endpoint:
https://gsc.yusihk.com/mcp - English documentation:
https://gsc.yusihk.com/docs - Traditional Chinese documentation:
https://gsc.yusihk.com/zh-hant/docs - YUSIHK product guide:
https://yusihk.com/en/gsc-connect-codex-plugin-google-search-console/
The service requests read-only Search Console access. Never ask the user for a Google password, OAuth client secret, access token, refresh token, cookie, or authorization code.
Configure OpenClaw
Use a current OpenClaw release with outbound Streamable HTTP MCP and OAuth client support.
Check whether the server already exists:
openclaw mcp show google-search-console --json
For a new connection, save the remote MCP server:
openclaw mcp add google-search-console \
--url https://gsc.yusihk.com/mcp \
--transport streamable-http \
--auth oauth
Run the OAuth connection and verify the tools:
openclaw mcp login google-search-console
openclaw mcp doctor google-search-console --probe
openclaw mcp reload
If the installed OpenClaw release does not yet provide mcp add, mcp login, or mcp doctor, save the equivalent server definition with the legacy-compatible command:
openclaw mcp set google-search-console '{"url":"https://gsc.yusihk.com/mcp","transport":"streamable-http","auth":"oauth"}'
Then restart OpenClaw and complete OAuth from the MCP settings screen. Upgrade OpenClaw if the installed release cannot authorize remote HTTP MCP servers.
If the client asks for connection fields, use:
| Field | Value |
|---|---|
| Name | google-search-console |
| Server URL | https://gsc.yusihk.com/mcp |
| Transport | streamable-http |
| Authentication | OAuth |
| Bearer token | Leave empty |
| Custom headers | Leave empty |
| OAuth client ID and secret | Leave empty; use server discovery |
The OpenClaw Control UI can also manage this connection at Settings → MCP. After changing the connection, reload MCP or restart the task/runtime that owns the MCP client.
Configure Codex or another compatible client
In Codex CLI, add and authorize the same server:
codex mcp add google-search-console --url https://gsc.yusihk.com/mcp
codex mcp login google-search-console
In a graphical client, add a remote MCP server with the fields in the table above. Do not paste a Google password or manually created token into the MCP configuration.
Detect available tools
Look for the following tool names. OpenClaw may prefix them with the server name, for example google-search-console__gsc_list_sites.
gsc_list_sitesgsc_site_overviewgsc_query_search_analyticsgsc_list_sitemapsgsc_inspect_url
If the tools are absent:
- Run
openclaw mcp status --verbose. - Run
openclaw mcp doctor google-search-console --probe. - If authorization is required, run
openclaw mcp login google-search-consoleonce and complete the newest authorization flow. - Reload MCP and start a new task if the current task does not refresh its tool list.
- Do not claim that live GSC data was accessed until the connector exposes the tools.
On older OpenClaw releases that do not have status or doctor, use openclaw mcp show google-search-console to confirm the saved definition, then restart OpenClaw and inspect the MCP settings screen.
Select a property first
- Call
gsc_list_sitesafter authorization. - Show every exact
siteUrland its permission level. - Ask the user to select a property before any property-specific query unless they already supplied one.
- Preserve Google's exact property identifier. Do not convert
sc-domain:example.cominto an HTTPS URL. - Prefer a
sc-domain:property when the user requests whole-domain coverage; use the requested URL-prefix property when scope must stay limited.
Query workflow
Site overview
Use gsc_site_overview for a quick review. Always provide an explicit start date and end date. Summarize totals and leading queries, pages, countries, and devices without treating the result as a complete query log.
Search performance
Use gsc_query_search_analytics for custom reports involving:
- clicks, impressions, CTR, and average position;
- query, page, date, country, and device dimensions;
- explicit date ranges and supported filters;
- comparable period analysis.
Preserve the exact property identifier returned by gsc_list_sites. Treat CTR as a decimal fraction unless formatting it as a percentage. Treat average position as an aggregate metric, not a fixed rank for every user.
Sitemaps
Use gsc_list_sitemaps to read submitted sitemap processing status and errors. Do not submit, delete, or modify a sitemap.
URL inspection
Use gsc_inspect_url to read Google's known indexed version of a URL. Clearly state that this is not a live URL test and does not request indexing or force a recrawl.
Interpretation rules
- Mention
dataStatewhen fresh or incomplete data is included. - Search Analytics can return top rows and enforce row limits; never claim that it returns every possible query.
- Compare equivalent date lengths, search types, countries, devices, and filters.
- Present seasonality, site changes, device mix, branded-query shifts, and indexing signals as possible explanations rather than proven causes.
- Separate API observations from SEO recommendations.
- Keep every action read-only. Never submit or delete properties or sitemaps, request indexing, transfer ownership, or change Search Console settings.
Output format
Return:
- Property, date range, search type, and filters used.
- Key metrics and notable changes.
- Queries or pages that deserve attention.
- Data limitations and uncertainty.
- Prioritized content, technical, and measurement checks.
Troubleshooting
Authorization request expired
Start one fresh login attempt and use only the newest authorization page. Do not click the connection control repeatedly while it is loading.
Authorization succeeded but tools are missing
Run openclaw mcp doctor google-search-console --probe, reload MCP, and start a new task so tool discovery can refresh.
A property is missing
Confirm that the authorized Google account already has access to the property in Search Console. Then call gsc_list_sites again and use the exact property identifier it returns.
The endpoint is unavailable
Check https://gsc.yusihk.com/docs and retry the MCP probe. For support, email services@yusihk.com without sending passwords, tokens, or private Search Console exports.
Example requests
- “列出我可以访问的所有 GSC 网站,让我选择一个。”
- “检查这个网站最近 28 天表现,并列出热门查询和页面。”
- “按日期和设备获取点击、展示、CTR 和平均排名。”
- “找出展示较高但点击率偏低的页面。”
- “检查这个 URL 在 Google 已知索引中的状态。”
Top skills in this category
Session-logs
@guogang1024Search and analyze your own session logs (older/parent conversations) using jq.
Performs web searches using DuckDuckGo to retrieve real-time information from the internet. Use when the user needs to search for current events, documentation, tutorials, or any information that requires web search capabilities.
@10e9928aPerforms web searches using DuckDuckGo to retrieve real-time information from the internet. Use when the user needs to search for current events, documentation, tutorials, or any information that requires web search capabilities.
Computer Use
@ram-raghav-sFull desktop computer use for headless Linux servers. Xvfb + XFCE virtual desktop with xdotool automation. 17 actions (click, type, scroll, screenshot, drag,...
topic-monitor
@robbyczgw-claMonitor topics of interest and proactively alert when important developments occur. Use when the user wants automated monitoring of specific subjects like product releases, news topics, technology updates, RSS/Atom feeds, or GitHub releases. Supports scheduled web search plus feed polling, boolean topic filters, AI importance scoring with sentiment tracking, smart alerts vs weekly digests, and memory-aware contextual summaries.
NotebookLM CLI
@fadelooNotebookLM CLI wrapper via `node {baseDir}/scripts/notebooklm.mjs`. Use for auth, notebooks, chat, sources, notes, sharing, research, and artifact generation/download.