geoseo-mcp
Unifies traditional SEO and Generative Engine Optimization (GEO) for Google, Bing, Yandex, and major LLMs, providing tools for search performance analysis, citation tracking, on-page audits, and internal link graph analysis.
README
geoseo-mcp
The only open-source MCP server that covers traditional SEO and Generative Engine Optimization (GEO) across Google, Bing, Yandex, and the major LLMs — in one server.
Why this exists
Every other SEO MCP picks one lane:
mcp-gsc— Google Search Console only.brightdata-mcp— web access with a GEO bolt-on, vendor-coupled.geo-optimizer-skill— GEO scoring only, no traditional SEO.- Frase / Conductor / Cairrot — closed SaaS.
geoseo-mcp unifies them. One MCP, one config, every search surface that matters in 2026:
| Surface | Tools |
|---|---|
| Google Search | GSC performance, URL inspection, sitemaps, Indexing API |
| Bing Webmaster | Verified-site stats, query/page stats, URL inspect, fast submit, crawl issues |
| Bing / Yandex / Naver / Seznam | IndexNow submission |
| Google AI Overviews | aio_check + aio_citation_check via SerpAPI |
| LLM citations | Perplexity, ChatGPT (web search), Claude (web search), Gemini (Google Search grounding), plus a multi_llm_* super-tool that fans out across all four in parallel |
| On-page | Title / meta / heading / schema / internal-link audit + 0-100 score |
| Internal link graph | Orphans, dead-ends, hubs, dangling links, TF-IDF related-page suggestions with anchor hints |
llms.txt |
Generate from a folder of HTML, validate against the spec |
| Trend tracking | Local SQLite snapshots of GSC + LLM citations + AIO; time-series trend tools |
MIT licensed. Runs locally over stdio. No hosting, no API key gating, your credentials never leave your machine.
Status
v0.3 — alpha. 36 tools across GSC, Bing Webmaster, IndexNow, ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, on-page + folder audits, llms.txt, internal link graph + suggestions, and local SQLite trend tracking. See ROADMAP.
Install
# Using uv (recommended)
uvx geoseo-mcp
# Or pip
pip install geoseo-mcp
Or from source:
git clone https://github.com/Rachit8484/geoseo-mcp.git
cd geoseo-mcp
pip install -e ".[dev]"
Try it without any credentials
Several tools work with zero configuration — useful to kick the tires before you wire up any API keys.
# audit a single page
echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"audit_page","arguments":{"source":"https://example.com"}}}' \
| uvx geoseo-mcp
# audit a whole folder of HTML
... "audit_site","arguments":{"folder":"./content"}
# build the internal link graph
... "internal_link_graph","arguments":{"folder":"./content","site_host":"yourdomain.com"}
# generate llms.txt
... "generate_llms_txt","arguments":{"folder":"./content","site_url":"https://yourdomain.com"}
In an MCP client (Cursor / Claude Desktop / Continue / Cline), it's even simpler — just ask:
"Audit my
./contentfolder and tell me the 5 worst-scoring pages" "Build the internal link graph and find orphan pages" "Generate an llms.txt for yourdomain.com from./content"
Configure your MCP client
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"geoseo": {
"command": "uvx",
"args": ["geoseo-mcp"],
"env": {
"GEOSEO_GOOGLE_CLIENT_SECRET": "/absolute/path/to/client_secret.json",
"GEOSEO_BING_WEBMASTER_API_KEY": "your-bing-key",
"GEOSEO_INDEXNOW_KEY": "your-indexnow-key",
"GEOSEO_PERPLEXITY_API_KEY": "pplx-...",
"GEOSEO_OPENAI_API_KEY": "sk-...",
"GEOSEO_ANTHROPIC_API_KEY": "sk-ant-...",
"GEOSEO_GEMINI_API_KEY": "AIza...",
"GEOSEO_SERPAPI_API_KEY": "..."
}
}
}
}
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the Windows equivalent:
{
"mcpServers": {
"geoseo": {
"command": "uvx",
"args": ["geoseo-mcp"],
"env": {
"GEOSEO_GOOGLE_CLIENT_SECRET": "/absolute/path/to/client_secret.json",
"GEOSEO_BING_WEBMASTER_API_KEY": "your-bing-key",
"GEOSEO_INDEXNOW_KEY": "your-indexnow-key",
"GEOSEO_PERPLEXITY_API_KEY": "pplx-...",
"GEOSEO_OPENAI_API_KEY": "sk-...",
"GEOSEO_ANTHROPIC_API_KEY": "sk-ant-...",
"GEOSEO_GEMINI_API_KEY": "AIza...",
"GEOSEO_SERPAPI_API_KEY": "..."
}
}
}
}
See examples/ for ready-to-paste configs.
Get credentials
- Google Search Console — see
docs/setup-gsc.md. One-time OAuth flow. - IndexNow — generate any random 32-char hex string and host it at
https://yourdomain.com/<key>.txt. Spec. - Perplexity — API key from settings. Pay-as-you-go, ~$1 per 1000 queries.
- OpenAI — API key. Uses the Responses API +
web_search_previewtool. - Anthropic — API key. Uses Claude with the
web_search_20250305server-side tool. - Gemini — API key from AI Studio. Uses Google Search grounding (the same signal Google's AI Overviews are built on).
- Bing Webmaster — API key from Bing Webmaster Tools → Settings → API access.
- SerpAPI — API key. Free tier: 100 searches/month. Used for Google AI Overviews tracking (
aio_check,aio_citation_check).
All credentials are optional — tools that need a key you don't have will return a clear error, the rest still work.
Tools (v0.3 — 36 total)
Status / discovery
geoseo_status— show which engines are configured.list_llm_engines— show which LLM engines have keys.
Google Search Console
gsc_list_sites,gsc_performance,gsc_inspect_url,gsc_submit_sitemap
Bing Webmaster Tools (new in v0.3)
bing_list_sites,bing_query_stats,bing_page_stats,bing_url_info,bing_submit_url,bing_crawl_issues
Indexing
indexnow_submit_url— single URL → Bing/Yandex/Naver/Seznam/Yep.indexnow_submit_urls— batch up to 10,000 URLs.
LLM citations (the GEO/AEO core)
perplexity_query/perplexity_citation_checkopenai_query— ChatGPT with web search.claude_query— Claude with server-side web search.gemini_query— Gemini with Google Search grounding.multi_llm_query— fan out one question to every configured LLM in parallel.multi_llm_citation_check— citation-share metrics for your domain across ChatGPT + Claude + Gemini + Perplexity in one call. This is the headline tool.
Google AI Overviews (new in v0.3)
aio_check— does AIO fire for this query? Which URLs does it cite?aio_citation_check— AIO fire-rate + citation-share for your domain across a batch of queries.
On-page audit
audit_page— local file or URL: title, meta, H1-H3, word count, schema, OG, freshness, GEO quotability, 0-100 score, list of findings.audit_site— recursive audit over a folder, with worst-pages report.
Internal link graph (new in v0.3)
internal_link_graph— orphans, dead-ends, hub pages, dangling hrefs.suggest_internal_links— TF-IDF related-page suggestions per file with anchor-text hints.
llms.txt
generate_llms_txt— produce a spec-compliantllms.txtfrom a folder of HTML pages, grouped by URL prefix.validate_llms_txt— validate an existing file/URL with line-numbered issue list.
Trend tracking (new in v0.3 — local SQLite)
trend_init,trend_statssnapshot_gsc— persist a GSC performance pull as a timestamped snapshot.snapshot_llm_citations— persistmulti_llm_citation_checkresults.snapshot_serp_aio— persist AIO citation results.trend_gsc— clicks/impressions/position over time (filterable by query/page).trend_llm_citations— citation share over time (per engine).
Snapshots are append-only and stored at $GEOSEO_DB (default: platform user-data dir / geoseo.sqlite). Schedule the snapshot_* tools weekly to build a private time-series of your AI-search visibility — no SaaS.
Roadmap
- v0.4 — Yandex Webmaster API, Grok citations, schema linter, broken-link prospector, vertical packs (YMYL / health / e-commerce).
- v1.0 — Stable API, full test coverage, optional remote (HTTP) transport.
See docs/architecture.md for the engine plug-in interface — adding a new search engine or LLM is one file.
Contributing
PRs welcome. Each engine lives in src/geoseo_mcp/engines/<name>.py and implements the Engine ABC. Add yours, register it in engines/__init__.py, expose tools in tools/, ship.
License
MIT. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。