SEOPro AI

io.github.alpha-ai-labs由 GitHub 帳號 alpha-ai-labs 發布v1.2.0更新於 Oct 24, 2026

Turn Search Console data into SEO actions, content, publishing, indexing, and AI insights.

已驗證Streamable HTTP可網頁執行效率與工作流資料與分析商業與電商

概覽

AI 產生的概覽

託管式 SEO MCP 伺服器，把 Search Console 資料轉化為稽核、關鍵字計畫、AI 文章生成、CMS 發佈與收錄提交。

可以做什麼
  • 列出帳戶下的網站，讀取帳戶脈絡、內容統計與自動營運狀態
  • 取得 Google Search Console 成效、熱門關鍵字、月度趨勢與 AI 助理導流資料
  • 對任意網址執行技術 SEO 稽核、檢查 AI 引用情形並檢視反向連結概況
  • 讀取與產生關鍵字／內容計畫，並啟動非同步文章生成、輪詢生成狀態
  • 列出與閱讀文章，將審核通過的草稿發佈到已連接的 CMS，並提交網址收錄及查詢收錄狀態
適用情境
適合管理一個或多個網站的創辦人、行銷人員、顧問與代理商，希望助理把 Search Console 洞察串成優先建議、內容生成、發佈與收錄的完整流程。
執行需求
遠端 Streamable HTTP 端點，不需本機執行環境或安裝套件。需要 SEOPro AI 帳戶（README 提到付費七天試用）、已連接的網站、搜尋成效流程所需的 Google Search Console，以及發佈流程所需的支援 CMS 連接。每個請求都需驗證：互動式用戶端使用 OAuth，開發者用戶端可用可撤銷的 SEOPro AI API 金鑰作為 bearer token。
安裝前請注意
發佈操作會把內容推送到線上 CMS（除非設定 publish_as_draft），收錄操作會向 Google／Bing 提交網址；兩者都需明確的確認參數，並應先取得使用者同意。產生關鍵字計畫時使用 replace_existing 會捨棄待處理主題。帳戶資料、網站內容與工具參數會傳送到 SEOPro AI 後端，後端會記錄營運中繼資料。API 金鑰應存放在用戶端密鑰儲存區，不要寫入設定檔。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 SEOPro AI，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。

Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.alpha-ai-labs%2Fseo-agent/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "seopro-ai": {
      "type": "http",
      "url": "https://api.seoproai.co/api/mcp"
    }
  }
}

README

SEOPro AI Agent Plugin

Run the full SEO loop from your AI assistant:

Search Console insight → prioritized action → generated article → CMS publishing → indexing → measurable Google and AI-referral results.

SEOPro AI is a hosted remote MCP server for anyone responsible for website growth, from founders and small teams to in-house marketers, consultants, and agencies. It is distributed as a portable Agent Plugin. The implementation stays on SEOPro AI infrastructure; this repository contains the public connection metadata, setup guides, reusable SEO workflows, and security documentation.

[Agency operators reviewing an SEO workflow]

Connect

Install as an Agent Plugin

This repository conforms to Agent Plugins 1.0.0. Add or clone the repository through a compatible client's plugin installation flow. The client discovers:

text
seoproai-mcp/├── plugin.json                   # Agent Plugins manifest├── mcp.json                      # Agent Plugins MCP config├── .claude-plugin/plugin.json    # Claude plugin manifest├── .mcp.json                     # Claude plugin MCP config└── skills/    └── seoproai-operator/        └── SKILL.md
  • plugin.json identifies the portable package.
  • mcp.json declares the hosted Streamable HTTP server.
  • .claude-plugin/plugin.json and .mcp.json make the same package native to Claude Code and Cowork.
  • skills/seoproai-operator/ supplies safe SEO workflows and approval gates for one or many websites.

Agent Plugins 1.0.0 intentionally leaves remote authentication to the client. Complete OAuth in the client or keep a scoped SEOPro AI API key in its secret store. Never add credentials to mcp.json, plugin.json, or SKILL.md.

Clients that do not yet load Agent Plugins can use the manual setup below.

Connect manually

Remote MCP URL

text
https://api.seoproai.co/api/mcp

No server package or local runtime is required. Interactive clients use OAuth. Developer clients may use a revocable SEOPro AI API key as a bearer token.

Official MCP Registry identifier: io.github.alpha-ai-labs/seo-agent

ClientAuthSetup
ClaudeOAuthSettings → Connectors → Add custom connector → paste the URL → Connect
ChatGPTOAuthSettings → Apps & Connectors → Advanced/Developer mode → Create → paste the URL
CodexOAuthcodex mcp add seopro-ai --url https://api.seoproai.co/api/mcp, then codex mcp login seopro-ai
Claude CodeOAuth or API keyclaude mcp add --transport http seopro-ai https://api.seoproai.co/api/mcp
HermesAPI keyAdd the URL and environment-backed bearer header in ~/.hermes/config.yaml
OpenClawOAuth or API keyAdd a Streamable HTTP entry under mcp.servers; prefer OAuth or keep a static key in an environment variable
Cursor / WindsurfOAuth or API keyAdd a remote Streamable HTTP server in MCP settings

See client setup for copy-paste configurations and current caveats.

Three SEO workflows

1. Find website quick wins

text
Using SEOPro AI, list my websites. For [website], review the last 28 days ofSearch Console performance. Find queries ranking 5–15 with meaningfulimpressions, group them by page, and give me a prioritized action plan. Do notmake any changes.

2. Build and generate a content plan

text
Using SEOPro AI, review [client site]'s search performance and current keywordplan. Propose a four-week content plan around [service/topic]. Show me the planbefore creating it. After I approve it, create the keyword plan and generateonly the first article.

3. Review, publish, index, and measure

text
Using SEOPro AI, show me the draft for [article]. Check its title, metadescription, structure, and relevance to the target keyword. Do not publishuntil I explicitly approve the final draft. After approval, publish it, submitthe published URL for indexing, and give me a 30-day measurement checklist.

The complete prompt pack is in agency prompt pack.

Demo and example data

[Example connected Search Console performance view]

The Search Console image is an example connected performance view. Results vary by website.

What the server exposes

The server provides 18 narrowly scoped tools across:

  • Websites and account context.
  • Search Console summaries, keywords, trends, and AI-referral traffic.
  • Technical audits, AI citations, and backlink profiles.
  • Keyword/content planning and asynchronous article generation.
  • Draft review and explicit-confirmation CMS publishing.
  • Explicit-confirmation indexing requests and indexing status.

Read tools are annotated as read-only and non-destructive. Publishing and indexing are annotated as externally consequential and require an explicit confirmation argument on every call.

Privacy and security

  • Authentication is required for every MCP request.
  • OAuth and API-key access can be revoked.
  • Website ownership and subscription limits are enforced by the same backend routes used by the SEOPro AI dashboard.
  • Analytics records operational metadata such as tool name, success, latency band, auth method, and normalized client family. It never records credentials, tool arguments, article content, website content, or returned customer data.
  • The public repository contains no hosted-server source code or credentials.

Read SECURITY.md, architecture and data handling, and the SEOPro AI Privacy Policy.

Requirements

  • An SEOPro AI account. The current acquisition offer is a $1 seven-day trial.
  • Connected websites and, for search-performance workflows, Google Search Console.
  • A supported CMS connection for publishing workflows.

Start at seoproai.co/mcp.

Repository map

  • plugin.json — Agent Plugins 1.0.0 package manifest.
  • mcp.json — Portable Streamable HTTP MCP connection without embedded credentials.
  • .claude-plugin/plugin.json — Claude Code and Cowork plugin manifest.
  • .mcp.json — Claude-native remote HTTP connector configuration without embedded credentials.
  • server.json — Official MCP Registry metadata for the hosted endpoint.
  • docs/client-setup.md — Client-specific setup.
  • docs/agency-prompt-pack.md — Safe, reusable agency prompts.
  • docs/compatibility.md — Tested client/auth matrix.
  • docs/troubleshooting.md — Connection and tool-call diagnostics.
  • docs/architecture-and-data-handling.md — Trust boundary and analytics disclosure.
  • docs/directory-listing-copy.md — Canonical short and long listing descriptions.
  • docs/90-second-demo-script.md — Timed narration, capture plan, and privacy checklist.
  • docs/example-tool-results.md — Fictional sample data for screenshots and reviews.
  • assets/ — Public workflow and product-proof visuals.
  • skills/seoproai-operator/ — Portable Agent Skill for founders, small teams, in-house marketers, consultants, and agencies.
  • scripts/validate_agent_plugin.py — Schema, transport, package-boundary, and skill validation used by CI.

Support

Email [email protected] or open a GitHub issue without including credentials or customer data. Service availability: seoproai.co.

License

Documentation and repository assets are licensed under MIT. The hosted SEOPro AI service and its implementation remain proprietary and are governed by the Terms.

來源：README.md，提交 b0ec8b7

工具

18

list_websites

List Account Websites

List the websites in the user's SEOPro AI account. Returns each website's numeric website_id (needed by most other tools), URL, platform and autopilot status. Call this first.

無需輸入參數。

get_website_stats

Get Website Content Stats

Get content stats for a website: total articles, keywords tracked, published posts and whether SEO autopilot is enabled.

輸入參數website_id必填

get_search_performance

Get Search Performance

Get the Google Search Console performance summary for a website: clicks, impressions, CTR, average position, top keywords, AI-assistant traffic sources and device/country breakdowns. Dates are YYYY-MM-DD; defaults to the last 30 days.

輸入參數end_datestart_datewebsite_id必填

get_top_search_keywords

Find Top Search Keywords

List the website's top search keywords with clicks, impressions, CTR and average position. Great for finding quick wins (positions 5-15 with high impressions). Dates are YYYY-MM-DD; defaults to the last 30 days.

輸入參數limitend_datestart_datewebsite_id必填

get_ai_search_traffic

Measure AI Search Traffic

Show traffic the website receives from AI assistants and AI search (ChatGPT, Perplexity, Claude, Gemini, etc.): sessions, engaged pages and per-source breakdowns. Dates are YYYY-MM-DD; defaults to the last 30 days.

輸入參數limitend_datestart_datewebsite_id必填

run_seo_audit

Run Technical SEO Audit

Run a technical SEO audit on any URL: meta tags, headings, images, links, performance, content quality, security and social metadata, with a scored result. Rate-limited for free accounts.

輸入參數url必填

check_ai_citations

Check AI Citations

Check whether AI assistants (ChatGPT, Claude, Gemini, Perplexity) mention or cite a website when answering questions in its niche. Takes 1-3 minutes; results are cached for 7 days.

輸入參數url必填

get_keyword_plan

Get Keyword Content Plan

Get the website's scheduled keyword/content plan: which keywords are planned, their scheduled dates and statuses.

輸入參數website_id必填

generate_keyword_plan

Generate Keyword Content Plan

Generate an AI keyword/content plan for the website. Optionally guide it with topic instructions and comma-separated primary/secondary keywords. Can take a minute or two. Set replace_existing=true only after the user explicitly approves discarding the current plan's pending topics; that path also requires confirm_replace_existing=true.

輸入參數website_id必填primary_keywordsreplace_existingsecondary_keywordstopic_instructionsconfirm_replace_existing

generate_article

Generate SEO Article

Start generating an SEO article for a target keyword. Returns immediately; generation takes 5-10 minutes in the background — poll get_article_status for the scheduled_date until it shows a blog. scheduled_date is YYYY-MM-DD and defaults to today (UTC). One article is generated per website per date; if that date already has an article the existing blog_id is returned. Counts toward the plan's monthly limit.

輸入參數topickeyword必填website_id必填scheduled_date

get_article_status

Check Article Generation Status

Check article generation/publication status per date. Dates are YYYY-MM-DD and default to the current month. A date with has_blog=true and a blog_id means the article is ready to read via get_article.

輸入參數end_datestart_datewebsite_id必填

list_articles

List Website Articles

List the website's articles (drafts and published), newest first. Optionally filter by status (e.g. 'draft', 'published').

輸入參數limitstatuswebsite_id必填

get_article

Review Article Draft

Read a single article by blog_id. By default returns metadata and a short excerpt; set include_content=true for the full HTML content.

輸入參數blog_id必填include_content

publish_article

Publish Article to CMS

Publish an article to the website's connected platform (WordPress, Shopify, Webflow, etc.). Set publish_as_draft=true to push it as an unpublished draft in the CMS instead of going live. Before calling, show the user the article and destination and obtain explicit approval. Every call requires confirm_publish=true.

輸入參數blog_id必填confirm_publishpublish_as_draft

request_indexing

Request Search Engine Indexing

Submit up to 50 URLs for search-engine indexing (Google/Bing) via SEOPro AI auto-indexing. Requires the auto-indexing feature on the account. Show the exact URL list and obtain explicit user approval first; every call requires confirm_indexing=true.

輸入參數urls必填website_id必填confirm_indexing

get_indexing_status

Get Indexing Status

Get auto-indexing stats for the website: how many URLs are indexed, pending or failed.

輸入參數website_id必填

版本歷史

1
  1. v1.2.0最新Sep 16, 2026

評分

已滿足適用檢查項的 100%

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 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分