Autorank Mcp

io.github.RyanKramer由 GitHub 帳號 RyanKramer 發布v0.1.0更新於 Oct 24, 2026

Technical SEO tools for AI agents: structured data, meta tags, robots, SERP truncation. No API key.

概覽

AI 產生的概覽

本機 MCP 伺服器,提供離線技術 SEO 工具,讓助理產生並驗證結構化資料、中繼標籤、robots 指令與搜尋結果預覽。

可以做什麼
為 FAQ、文章、麵包屑與在地商家頁面產生 JSON-LD 結構化資料,並驗證既有 JSON-LD 區塊的必填欄位與風險寫法。產生完整中繼標籤 head 區塊、robots meta 標籤、robots.txt 與 hreflang 標籤,並驗證 robots.txt 語法。依像素寬度檢查標題與描述是否被截斷,預覽搜尋結果的整體呈現,並評估關鍵字密度與可讀性。產生帶 UTM 參數的連結與乾淨的網址 slug,並列出需要即時爬取的網頁版工具。
適用情境
適合在撰寫或稽核網站頁面 SEO 時使用:產生 schema 標記、head 標籤與爬取指令,或檢查標題與描述在搜尋結果中如何被截斷。全部離線執行,適合在 MCP 用戶端內做快速、可重複的檢查。
執行需求
以本機程序透過 stdio 執行,從 PyPI 套件 autorank-mcp 安裝,例如使用 uvx 或 pip。不需要 API 金鑰、帳號、環境變數或網路存取。僅支援桌面用戶端。
安裝前請注意
工具皆為純函式,只回傳產生的標記與報告,不會代你寫入、傳送或發布任何內容。產生的 schema 與 robots 指令在部署前仍需人工複核,因為與頁面內容不一致的結構化資料會被視為垃圾內容訊號。需要即時爬取的部分 SEO 檢查並未包含在內。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Autorank Mcp,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

AutoRank MCP

Technical SEO tools for AI agents. Generate and validate structured data, build meta tags and crawl directives, check how a result will actually render in Google, and analyse copy — all from inside Claude, Cursor, or any MCP client.

No API key. No account. No network calls. Every tool is a pure function, so the server does its job offline and installs in one line.

Browser versions of these tools live at autorank.so/free-tools.

Install

Claude Desktop — add to claude_desktop_config.json:

json
{  "mcpServers": {    "autorank": {      "command": "uvx",      "args": ["autorank-mcp"]    }  }}

Or with pip:

bash
pip install autorank-mcp

Then point your client at the autorank-mcp command. No environment variables are required.

Tools

Structured data

ToolWhat it does
generate_faq_schemaFAQPage JSON-LD from question/answer pairs
generate_article_schemaArticle JSON-LD with author and dates
generate_breadcrumb_schemaBreadcrumbList JSON-LD
generate_local_business_schemaLocalBusiness JSON-LD with postal address
validate_json_ldParses a block, checks required fields per type, flags risky patterns

Meta and crawl directives

ToolWhat it does
generate_meta_tagsFull head block: title, description, canonical, OG, Twitter Card
generate_robots_metaRobots meta tag from directives, with a plain-English explanation
generate_robots_txtrobots.txt, optionally blocking the major AI crawlers
validate_robots_txtValidates syntax and surfaces the mistakes that cause real damage
generate_hreflanghreflang tags, with the return-link requirement spelled out

SERP and copy

ToolWhat it does
check_title_lengthPixel-width truncation check for titles
check_description_lengthPixel-width truncation check for descriptions
serp_previewHow the whole result renders, breadcrumb included
keyword_densityWord and phrase frequency with a stuffing warning
readability_scoreFlesch Reading Ease and grade level

Utilities

ToolWhat it does
build_utm_linkUTM-tagged tracking URLs, warns on case-inconsistency
slugifyClean URL slugs from titles
autorank_free_toolsLists the web-only tools that need live crawling

Design notes

Title checks measure pixels, not characters. Google truncates on rendered width, and character counts mislead badly — sixty capital Ws and sixty is occupy very different space. Titles are the most common cause of a rewritten snippet, so the check approximates width instead.

The schema tools will not help you lie. validate_json_ld flags aggregateRating whenever it appears, because fabricated ratings are a documented cause of manual actions, and generate_faq_schema states plainly that the answers must already be visible on the page. Structured data that disagrees with rendered content is a spam signal, not a shortcut.

Some tools are deliberately absent. Anything needing live crawling or a keyword corpus — broken-link checking, keyword difficulty, sitemap extraction, duplicate-content detection — is not a pure function, so it stays on the web where it belongs. autorank_free_tools lists those.

robots.txt controls crawling, not indexing. validate_robots_txt says so in its output, because conflating the two is the single most common way sites accidentally leave pages in the index.

Development

bash
git clone https://github.com/RyanKramer/autorank-mcpcd autorank-mcppip install -e .autorank-mcp

About

Built by AutoRank — AI SEO content automation. The web versions of these tools, plus about thirty more, are free at autorank.so/free-tools.

MIT licensed.

來源:README.md,提交 fa8ad4c

工具

0
SourceWeft 尚未在沙箱中啟動過該軟體包,因此其工具未知。

版本歷史

1
  1. v0.1.0最新Sep 16, 2026

評分

已滿足適用檢查項的 50%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通25 分

    請保持端點能夠回應 MCP 握手;SourceWeft 每 30 天探測一次。

  • 已宣告認證方式10 分

    請在註冊中心條目中宣告憑證請求標頭,或保持端點能夠回應探測,以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分