io.github.saurabhsharma2u/search-console-mcp
搜索与获取by saurabhsharma2u
面向 Google Search Console 的 MCP server,支持站点、sitemaps、分析数据与 URL 检查。
什么是 io.github.saurabhsharma2u/search-console-mcp?
面向 Google Search Console 的 MCP server,支持站点、sitemaps、分析数据与 URL 检查。
README
🔍 Search Console MCP
Google Search Console + Bing Webmaster Tools + GA4 — in one context window.
Stop exporting CSVs. Start asking your AI agent questions.
📚 Docs · Quick Start · Tools · Security
</div>Why this exists
SEO data lives in three different silos. Answering one question — "did my traffic drop because of a ranking loss or a UX issue?" — usually means logging into three dashboards, exporting three CSVs, and doing VLOOKUPs by hand.
Search Console MCP puts GSC, Bing, and GA4 behind one set of tools your AI agent can call directly, and does the SEO math (cannibalization, anomaly detection, opportunity scoring) before the data ever reaches your context window — so your agent gets insights, not spreadsheets.
| Before | After | |
|---|---|---|
| Data | 3 dashboards, manual exports | 1 unified context |
| Analysis | Manual VLOOKUPs & pivot tables | Deterministic SEO math, done server-side |
| Accounts | Constant re-login | 20+ accounts, auto-resolved per site |
| Insight | Raw rows, agent guesses | Curated signals (opportunity scores, anomalies) |
⚡ Quick Start
npx search-console-mcp setup
This opens your browser, authorizes your Google account, and stores your credentials securely (see Security). Then add it to your MCP client config (Claude Desktop, Cursor, etc.):
{
"mcpServers": {
"search-console": {
"command": "npx",
"args": ["search-console-mcp"]
}
}
}
Restart your client — and try one of the prompts below.
💬 Try it
Paste these straight into your agent:
"My traffic dropped this week vs. last. Find exactly when it started and which pages are responsible."
"Find keywords for example.com ranking positions 8–15 with 1,000+ impressions — my best quick wins."
"Check for keyword cannibalization — are two of my pages competing for the same query?"
<details> <summary>More example prompts</summary>"Run
opportunity_matrixon my top 20 pages: which have high search visibility but poor on-site engagement?"
- "Run a full SEO health check, segmented by Brand vs Non-Brand, with 3 high-impact actions."
- "Fetch my top 5 pages by impressions and run a PageSpeed audit — any correlation with declining rankings?"
- "Compare Google vs Bing performance for the last 30 days — where is Bing winning?"
- "Am I too dependent on Google? Flag keywords where 85%+ of clicks come from one engine."
🔌 Connect your accounts
| Platform | Method | Setup |
|---|---|---|
| Google Search Console | OAuth (recommended) | npx search-console-mcp setup |
| Google Search Console | Service Account | Set GOOGLE_APPLICATION_CREDENTIALS — details |
| Bing Webmaster Tools | API Key | export BING_API_KEY="..." — get a key |
| Google Analytics 4 | Service Account | npx search-console-mcp setup --engine=ga4 |
Manage everything from the CLI:
npx search-console-mcp accounts list
npx search-console-mcp accounts add-site --account=you@company.com --site=example.com
npx search-console-mcp accounts remove --account=you@company.com
When your agent queries a site, the server auto-resolves which account owns it — no manual switching. Multi-account docs →
🖥️ Run tools from the CLI
Search Console MCP also exposes registered MCP tools as direct CLI commands. Use the run subcommand to list tools, inspect tool-specific arguments, and print results as JSON, CSV, or an ASCII table:
# List registered tools
npx search-console-mcp run --help
# Show options for one tool
npx search-console-mcp run analytics_query --help
# Run a tool with JSON output (default)
npx search-console-mcp run seo_low_hanging_fruit --siteUrl=https://example.com --minImpressions=100
# Print array results as CSV or a table
npx search-console-mcp run analytics_query --siteUrl=https://example.com --startDate=2026-06-01 --endDate=2026-06-30 --dimensions=date,query --format=csv
npx search-console-mcp run sites_list --engine=google --format=table
CLI arguments are validated against the same Zod schemas used by MCP clients. String arguments are coerced for common schema types, including numbers, booleans, comma-separated arrays, and JSON arrays/objects.
<details> <summary id="service-account-advanced">Service Account setup (for servers/automation)</summary>- Create a service account in the Google Cloud Console
- Generate a JSON key
- Add the service account email as a user in Search Console with "Full" or "Restricted" access
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/key.json"
🛠 Tools
40+ tools across five categories. A few flagship ones:
| Tool | What it does |
|---|---|
opportunity_matrix | Ranks pages by combining GSC visibility with GA4 engagement — where's the ROI? |
seo_striking_distance | Keywords ranking 8–15 — your fastest wins |
seo_cannibalization | Pages competing for the same query, with a recommended primary |
analytics_anomalies | Statistical traffic drop/spike detection |
sites_health_check | One-shot WoW performance + sitemap + anomaly check |
compare_engines | Google vs Bing performance, side by side |
Analytics & Trends
analytics_query · analytics_trends · analytics_anomalies · analytics_drop_attribution · analytics_time_series · analytics_compare_periods · seo_brand_vs_nonbrand
SEO Opportunities
seo_low_hanging_fruit · seo_striking_distance · seo_low_ctr_opportunities · seo_cannibalization · seo_lost_queries
SEO Primitives (building blocks for agent logic)
seo_primitive_ranking_bucket · seo_primitive_traffic_delta · seo_primitive_is_brand · seo_primitive_is_cannibalized
Sites & Sitemaps
sites_list · sites_add · sites_delete · sites_health_check · sitemaps_list · sitemaps_submit
Inspection & Validation
inspection_inspect · pagespeed_analyze · schema_validate
URL Indexing
indexing_submit_url · indexing_remove_url · indexing_status · indexing_batch_submit · bing_url_submission_quota
Bing Webmaster Tools
bing_sites_list · bing_analytics_query · bing_opportunity_finder · bing_seo_recommendations · bing_url_info · bing_index_now · bing_crawl_issues · bing_analytics_detect_anomalies · bing_analytics_time_series · bing_seo_lost_queries · bing_brand_analysis · bing_sitemaps_list · bing_sitemaps_submit
Google Analytics 4
analytics_page_performance · analytics_traffic_sources · analytics_organic_landing_pages · analytics_content_performance · analytics_conversion_funnel · analytics_user_behavior · analytics_audience_segments · analytics_realtime · analytics_ecommerce · analytics_pagespeed_correlation
Cross-Platform Intelligence
opportunity_matrix · page_analysis · traffic_health_check · brand_analysis · compare_engines
🔒 Security
- OS keychain first — tokens stored in macOS Keychain, Windows Credential Manager, or Linux Secret Service
- AES-256-GCM fallback — encrypted with a key derived from your machine's hardware ID; a stolen file is useless on another device
- Minimal storage — only
refresh_tokenandexpiry_dateare persisted, atmode 600 - Silent refresh — tokens renew automatically in the background
License
<div align="center">If this saves you a spreadsheet, consider ⭐ starring the repo.
</div>常见问题
io.github.saurabhsharma2u/search-console-mcp 是什么?
面向 Google Search Console 的 MCP server,支持站点、sitemaps、分析数据与 URL 检查。
相关 MCP Server
Brave 搜索
编辑精选by Anthropic
Brave Search 是让 Claude 直接调用 Brave 搜索 API 获取实时网络信息的 MCP 服务器。
✎ 如果你想让 AI 助手帮你搜索最新资讯或技术文档,这个工具能绕过传统搜索的限制,直接返回结构化数据。特别适合需要实时信息的开发者,比如查 API 更新或竞品动态。不过它依赖 Brave 的 API 配额,高频使用可能受限。
Puppeteer 浏览器控制
编辑精选by Anthropic
Puppeteer 是让 Claude 自动操作浏览器进行网页抓取和测试的 MCP 服务器。
✎ 这个服务器解决了手动编写 Puppeteer 脚本的繁琐问题,适合需要自动化网页交互的开发者,比如抓取动态内容或做端到端测试。不过,作为参考实现,它可能缺少生产级的安全防护,建议在可控环境中使用。
网页抓取
编辑精选by Anthropic
Fetch 是 MCP 官方参考服务器,让 AI 能抓取网页并转为 Markdown 格式。
✎ 这个服务器解决了 AI 直接处理网页内容时格式混乱的问题,适合需要让 Claude 分析在线文档或新闻的开发者。不过作为参考实现,它缺乏生产级的安全配置,你得自己处理反爬虫和隐私风险。