AEO Backlink

dev.advancelabs由 advancelabs.dev 發布官方發布者v1.0.1更新於 Oct 31, 2026

Find brand mentions, link prospects and competitor link sources, and draft outreach. Read-only.

已驗證Streamable HTTP可網頁執行網頁搜尋與抓取商業與電商

概覽

AI 產生的概覽

託管型 MCP 伺服器,用於反向連結開發研究:找出品牌提及、連結機會與競品連結來源,驗證連結並草擬開發信。

可以做什麼
  • 找出網路上對品牌的已連結與未連結提及,並依是否位於品牌網域分類
  • 依主題找出客座文章、資源頁與彙整頁類的連結開發機會,並依主機合併結果
  • 以免費搜尋訊號近似找出連向競品的頁面,並排除競品自身網域
  • 抓取頁面確認是否含有指向目標網域的連結,並回報錨點文字與 rel 屬性
  • 從頁面擷取聯絡信箱與社群帳號,並查詢某網址的 Wayback Machine 快照時間軸
  • 使用自己的 LLM 金鑰草擬含主旨與內文的個人化開發信
適用情境
適合 SEO、AEO 或反向連結工作:讓助理蒐集提及與潛在對象清單、確認頁面是否真的連到某網站,並在沒有付費反向連結索引的情況下準備開發信草稿。
執行需求
遠端 Streamable-HTTP 端點,無需在本機安裝。連線不需要任何憑證。開發信工具需按請求提供自己的 LLM API 金鑰。需要能連上該託管服務,該服務會再查詢 DuckDuckGo、目標頁面與 Wayback Machine。
安裝前請注意
開發信工具需按請求傳入你自己的 LLM API 金鑰;供應方稱僅在請求範圍內使用、不會保存或記錄,但它仍是交給第三方的密鑰。競品連結資料明確只是近似結果,並非完整反向連結索引。伺服器會抓取你指定的任意頁面。工具被描述為唯讀,但郵件工具會產生你可能送出的內容。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 AEO Backlink,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/dev.advancelabs%2Faeo-backlink/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": {
    "aeo-backlink": {
      "type": "http",
      "url": "https://aeo.advancelabs.dev/api/mcp/backlink/mcp"
    }
  }
}

README

Open-source TypeScript packages for Answer Engine Optimization (AEO), Generative Engine Optimization (GEO), and AI citation visibility. Contributions welcome, see CONTRIBUTING.md.

Open core. This repo is the open-source half of the AEO Toolkit: the audit engine, crawler, parser, schema validator and the Chrome extension, all Apache-2.0 and published to npm. The hosted platform behind advancelabs.dev/tools and the three MCP servers (the console, OAuth, billing, the Google and Bing integrations) moved to a private repository on 2026-10-20. Code published here before that date stays Apache-2.0.

Try it without installing anything: the six browser tools run free in the browser at advancelabs.dev/tools, with no sign-up and no account. Point the auditor at a URL and it returns a weighted, per-rule report in about a minute. Want it done for you? Advance Labs runs a monthly AEO retainer.

The hosted side is six browser tools (technical audit, E-E-A-T scan, llms.txt generator, GA4 + GSC chat, backlink graph, authority) and three MCP servers (ai-visibility, backlink, search) exposing 32 tools to Claude or any MCP client. They are proprietary and run on advancelabs.dev. This repo is the open core: the audit engine packages, the Chrome extension and the skills. See docs/reference/tools.md for the full map.


What it looks like

The hosted site is drawn in dithered pixel art: a figure pushing a faceted sphere up a slope under a dusk sky.

[The MCP page: Connect the AEO Toolkit to your AI client]

What is AEO?

Answer Engine Optimization (AEO) is the practice of structuring your content so that AI assistants (ChatGPT, Perplexity, Claude, Gemini) cite your site when answering questions in your domain. Traditional SEO gets you ranked on the blue-link results page. AEO gets you quoted in the AI answer.

As AI-powered search becomes the default discovery layer, AEO is the new SEO.


The audit

The hosted technical audit runs 44 checks (31 technical SEO + 13 AEO) across crawlability, AI-bot access, structured data, metadata and answer readiness, scored out of 100, with a prioritised fix list and templates for any crawl-hint file you're missing. The E-E-A-T scan is a separate tool that scores 14 signals. The engine defines 58 in total (44 + 14), but the audit itself runs 44. Free, no account, and your data stays yours.


The Chrome extension

AEO/GEO Auditor puts the same engine on your toolbar. Open any page, click the icon, and get a 0 to 100 AI-readiness score with a letter grade and a list of what to fix. Export it as a PDF.

It runs auditRules from packages/scoring/src/audit.ts: 44 checks across 9 categories (31 technical-SEO + 13 AEO) from extension 0.3.0, built on scoring 0.3.x. The Chrome Web Store build before that (0.2.0) ran 40, so the listing changes when 0.3.0 is published. The 14 E-E-A-T signals are a separate scorer the extension does not run, 58 is the engine total (44 + 14), not what any one audit runs.

Everything happens in your browser. No account, no server, no analytics, no telemetry. The only network requests are to the audited site's own robots.txt, sitemap.xml, and llms.txt, which is what the host permission is for. Nothing is stored, because nothing is sent.

Build it yourself:

bash
pnpm --filter @advance-labs/chrome-extension package# → apps/chrome-extension/aeo-extension.zip, loadable via chrome://extensions (Developer mode → Load unpacked, on dist/)

Packages

PackageDescription
@advance-labs/crawlerMulti-threaded web crawler with robots.txt compliance, sitemap parsing, and per-host rate limiting
@advance-labs/html-parserExtracts meta tags, Open Graph, Twitter Cards, headings, images, links, and JSON-LD from HTML
@advance-labs/schema-validatorValidates Schema.org JSON-LD structured data against known types
@advance-labs/scoringScores pages on 44 technical SEO and AEO checks, plus a separate 14-signal E-E-A-T scorer, with per-rule explanations

Plus @advance-labs/net-guard and @advance-labs/types. The one application here is apps/chrome-extension; the hosted console is not open source.


Quick Start

Six packages are published to npm and usable on their own:

bash
npm i @advance-labs/scoring      # the 44-check audit enginenpm i @advance-labs/net-guard    # SSRF-safe fetch: re-checks every redirect hopnpm i @advance-labs/crawler      # polite crawler with robots.txt + rate limitingnpm i @advance-labs/html-parsernpm i @advance-labs/schema-validatornpm i @advance-labs/types        # shared types, a dependency of the above

The hosted tools at advancelabs.dev/tools need no install at all.

bash
git clone https://github.com/Advance-Labs/aeo-toolkit.gitcd aeo-toolkitpnpm installpnpm build          # turbo builds every packagepnpm test           # the full suite, no network

Read the full documentation at docs.advancelabs.dev/aeo-toolkit.

Once built, the packages compose like this:

js
import { crawl } from '@advance-labs/crawler'import { parseHtml } from '@advance-labs/html-parser'import { analyzeStructuredData } from '@advance-labs/schema-validator'import { buildAuditReport } from '@advance-labs/scoring'
const result = await crawl('https://example.com/', { maxPages: 2 })const html = result.pages.filter((p) => p.ok && p.body && p.contentType?.includes('html'))
const report = await buildAuditReport(  {    crawl: result,    pages: html.map((p) => parseHtml(p.body, p.finalUrl)),    structuredData: html.map((p) => analyzeStructuredData(p.body, p.finalUrl)),    mode: 'full-site',  },  { durationMs: 0, version: '1.0.0' },)
console.log(report.url, report.score.overall, report.score.grade)console.log(report.topFixes.slice(0, 3).map((f) => f.title))

Connect the MCP servers

[mcptry]

Listed in the official MCP Registry as dev.advancelabs/aeo-search, dev.advancelabs/aeo-visibility and dev.advancelabs/aeo-backlink.

The three MCP servers are hosted, not installed. They are Streamable-HTTP endpoints, so there is nothing to npm install and no local process to run. Point any MCP client at the URL.

Claude Code: one command per server:

bash
claude mcp add --transport http --scope user aeo-visibility https://aeo.advancelabs.dev/api/mcp/ai-visibility/mcpclaude mcp add --transport http --scope user aeo-backlink   https://aeo.advancelabs.dev/api/mcp/backlink/mcpclaude mcp add --transport http --scope user aeo-search     https://aeo.advancelabs.dev/api/mcp/search/mcp

Drop --scope user to register them in the current project only. Check the result with claude mcp list.

Claude.ai / Claude Desktop: Settings → Connectors → Add custom connector, then paste the URL.

Cursor (~/.cursor/mcp.json), Windsurf, or any client that takes a JSON block:

json
{  "mcpServers": {    "aeo-visibility": { "url": "https://aeo.advancelabs.dev/api/mcp/ai-visibility/mcp" },    "aeo-backlink":   { "url": "https://aeo.advancelabs.dev/api/mcp/backlink/mcp" },    "aeo-search":     { "url": "https://aeo.advancelabs.dev/api/mcp/search/mcp" }  }}
ServerToolsCredentials
aeo-visibility5None to connect. Citation checks take a Perplexity key per request.
aeo-backlink7None.
aeo-search20Sign in with Google: the client opens a browser on first use (in Claude Code, /mcp → Authenticate). Or BYOK: a Google access token as Authorization: Bearer, plus an optional x-bing-api-key.

Three things that trip people up:

  • The trailing /mcp is required. The bare /api/mcp/<slug> returns the adapter's own "Not found", which looks like a routing bug and is not.
  • Only aeo-search has a login, and its discovery is path-scoped. It answers an anonymous request with a 401 pointing at /.well-known/oauth-protected-resource/api/mcp/search/mcp; its authorization server is /api/mcp/oauth. The ROOT /.well-known/oauth-* documents deliberately return 404, because aeo-visibility and aeo-backlink need no login.
  • Every tool is read-only. None calls a write method on any upstream API.

/api/mcp/ga-gsc/mcp still works as a compatibility alias for aeo-search; new configs should use /api/mcp/search/mcp.


Why AEO?

When a user asks ChatGPT "what is the best tool for X?", the answer comes from indexed content that AI models trust, not from ad-auction bidding. The trust signals for AI citation are:

  • Structured data (JSON-LD Organization, FAQPage, HowTo)
  • llms.txt: the AI equivalent of robots.txt
  • Canonical, server-rendered HTML: AI crawlers prefer static content
  • Topical authority: consistent, in-depth coverage of a topic

AEO Toolkit automates auditing all of these.


Monorepo Structure

aeo-toolkit/├── packages/          # the six published libraries, plus shared config├── apps/chrome-extension/├── apps/docs/         # Astro + Starlight docs site (renders ../../docs)├── docs/              # the documentation itself└── brand/             # the mark, lockups, palette and type (brand/README.md)

Built with Turborepo · TypeScript 5 · Vitest · React 19


Design

The site is pixel art: a figure pushing a faceted sphere under a dithered dusk sky. Art is drawn at a coarse cell, snapped to a small palette (ink, paper, two dusk oranges) with a 4x4 Bayer ordered dither and scaled up with hard edges. Labels use Geist Pixel Square, headings and body Geist Sans, code Geist Mono. Corners are square and shadows are hard-offset. The palette, type, tokens and rules are in brand/README.md, with the assets beside it. Anything a user sees should follow it, and docs/CONVENTIONS.md says how in code.


Made by Advance Labs

Built and maintained by Advance Labs · All Advance Labs products

AEO Toolkit is built and maintained by Advance Labs Inc., a software studio building Creatin, Cartrix, and this toolkit.

This project dogfoods its own tooling: advancelabs.dev ships with llms.txt, JSON-LD Organization schema, SSG-rendered pages, and canonical URLs, all patterns the scoring engine teaches.


Code of Conduct

We follow the Contributor Covenant: be respectful and constructive in all project spaces, and report unacceptable behavior to [email protected].


© 2026 Advance Labs Inc. · advancelabs.dev

來源:README.md,提交 8381d1a

工具

7

find_mentions

Find brand mentions

Find linked and unlinked mentions of a brand across the web via DuckDuckGo HTML results. When a domain is supplied, classifies each mention as linked (on the brand domain) or unlinked (a third-party page — a potential backlink opportunity).

輸入參數brand必填limitdomain

find_prospects

Find link prospects

Find guest-post, resource-page, and roundup link-building prospects for a topic by running classic outreach footprints through DuckDuckGo and merging the results by host.

輸入參數topic必填limitPerAngle

extract_contact_info

Extract contact info

Fetch a page and extract contact emails (including de-obfuscated and mailto links) and social media handles (Twitter/X, LinkedIn, Facebook, Instagram, GitHub, YouTube).

輸入參數url必填

check_page_history

Check page history

Return a Wayback Machine timeline for a URL: total snapshots, first/last archived dates, and a bounded list of snapshot records with direct archive links.

輸入參數url必填limit

generate_outreach_email

Generate outreach email

Draft a short, personalized cold outreach email from a contact + context object using your own LLM key (BYOK). Returns a subject line and body. The API key is request-scoped and is never persisted or logged.

輸入參數model必填apiKey必填contact必填context必填provider必填maxTokenstemperature

版本歷史

2
  1. v1.0.1最新Oct 31, 2026
  2. v1.0.0Oct 31, 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

其他 Advancelabs 整合