GEO and SEO Readiness MCP Server 2.1
Playwright GEO y auditoría SEO MCP con puntuación 5D y PDFs A4 white-label de 4 páginas. GEO and SEO Readiness MCP Server 2.1 es un servidor Model Context Protocol (MCP) listo para producción para desarrolladores indie, agencias SEO y creadores de contenido que pierden tráfico orgánico frente a los motores de respuestas de IA. Ejecuta renderizado dinámico con Playwright headless y evalúa cualquier URL mediante una matriz de puntuación de 5 dimensiones, genera llms.txt y esquema JSON-LD, compara citas de IA de la competencia y exporta un informe PDF A4 white-label de 4 páginas listo para el cliente. Solo License $49 pago único / Agency White-Label $129 pago único: https://dejunchen770231.gumroad.com/l/geo-seo-mcp-pro
Documentación
$49
Try before you buy: the free, MIT-licensed GEO & SEO Readiness MCP Lite runs the core checks (AI crawler access in robots.txt, llms.txt, JSON-LD and canonical tags, sitemap) with zero dependencies. This paid edition adds SPA rendering, weighted scoring, competitor comparison, candidate patches and white-label PDF reports.
Turn AI Search Audits Into Clear Client Deliverables
Audit real websites, benchmark competitors, generate candidate remediation files, and export branded reports from your MCP client.
Downloadable local software. Requires Python 3.11 or 3.12, an MCP client with stdio support, and installation of Python dependencies and Chromium. AI client subscriptions, hosting, and done-for-you consulting are not included.
Choose Your License
Solo Consultant Edition — $49 USD: one seat for personal and freelance client audits, full source, all four tools, patches, and bilingual PDF exports.
Agency White-Label Edition — $129 USD: team use within your agency, unlimited client pitch PDFs, custom agency branding/contact details and AgencyPackage pricing table, and full source.
Both editions contain the same complete software ZIP. The difference is the permitted user scope: Solo is single-user; Agency covers your agency team. Prices are one-time USD amounts before applicable taxes.
GEO and SEO Readiness MCP Server 2.1 (Commercial Agency Edition) is a local Python Model Context Protocol (stdio) server engineered for SEO agencies, technical marketing consultants, and growth engineers. Audit outputs retain evidence references and report exports include an audit sidecar and PDF SHA-256 hash.
12 Commercial Capabilities
Target Scope: Audits root domains and specific landing pages (preserves path and query parameters)
Dynamic SPA & JS Frameworks: Bounded 2-second headless Chromium DOM observation (render_mode=auto/always/never)
AI Crawler Classification: Separates live AI Search Bots (OAI-SearchBot, PerplexityBot, Claude-SearchBot) from Training Bots (GPTBot, ClaudeBot, Google-Extended)
Blocked or Unmeasured Data: Marks blocked/incomplete dimensions as unmeasured and recalculates measured_weight transparently
Competitor Benchmarking: Runs two real audits, checks scope/route comparability, and computes a signed score_gap (client - competitor)
Large XML Sitemaps: Bounded 4 MiB streaming parser + gzip support (extracts up to 2,000 closed <loc> entries safely via defusedxml)
Network & Proxy Routing: Configurable GEO_PROXY_MODE=auto/fallback/direct with HTTP_PROXY, HTTPS_PROXY, and ALL_PROXY support
Commercial Value Framing: Optional transparent upper-bound scenario math (monthly_visitors * avg_deal_usd * 0.03 * gap_ratio), only when you supply both inputs
robots.txt Patch Safety: Full-preservation candidate file; refuses automatic overwrite if the source robots.txt was truncated or unparseable
Client Deliverable Output: 4-Page White-Label A4 PDF Report (English & Simplified Chinese templates) +.audit.json evidence sidecar
Agency Pricing Customization: Injects your agency_name, agency_contact, and custom AgencyPackage tier table directly into Page 4 of the PDF
Verification & Test Suite: Regression tests and MCP stdio protocol checks included in the package.
4 Asynchronous MCP Tools Included Out of the Box
1. audit_website_geo_readiness
Audits a root domain or a specific commercial landing page across the 5-dimension readiness-1.1 weighted scoring model:
Search Crawler Access (30% Weight): Parses robots.txt against every sampled URL using urllib.robotparser. Evaluates OAI-SearchBot, PerplexityBot, and Claude-SearchBot independently from AI training crawlers (GPTBot, ClaudeBot, Google-Extended). Blocking training crawlers never penalizes a client's AI search score.
JSON-LD Structured Data (25% Weight): Extracts <script type="application/ld+json"> across sampled pages, recursively resolving nested arrays and @graph nodes. When render_mode="auto" detects a sparse JavaScript shell, the isolated headless Chromium runtime observes the live DOM for 2 seconds to capture client-rendered schemas.
LLMs Readability File llms.txt (15% Weight): Verifies /llms.txt content type and structure (H1 title = 40 pts, H2 sections = 30 pts, valid HTTP/HTTPS Markdown links = 30 pts) while rejecting HTML soft-404 pages.
XML Sitemap Integrity (15% Weight): Validates robots.txt Sitemap declarations or /sitemap.xml (urlset and sitemapindex), supporting compressed.xml.gz files and streaming up to 4 MiB without memory spikes.
Response Speed / TTFB (15% Weight): Measures real Time-To-First-Byte (<= 800ms = 100 pts, <= 1800ms = 70 pts, <= 3000ms = 40 pts, > 3000ms = 10 pts) and records the exact network_route (direct vs. proxy).
2. compare_competitors_geo
Loads or executes two real website audits and builds a 5-dimension comparison matrix with direct evidence references (<audit_id>.dimensions.<key>.score). It verifies whether both sites share the same scoring_version, audit_scope, measured_weight coverage, and network_route before computing the overall score_gap (client_score - competitor_score).
3. generate_client_pitch_and_patches
Transforms technical audit snapshots into an executive diagnostic pitch and three deployment-ready candidate files saved under snapshots/patches/<audit_id>/:
robots.txt Candidate: Preserves every existing rule and sitemap declaration in your client's live robots.txt and appends missing AI search crawler permissions at the bottom. If the live robots.txt was truncated or malformed, it halts automatic generation and flags manual review.
llms.txt Draft: Built exclusively from real audited page titles, meta descriptions, headings, and canonical URLs discovered during the crawl.
schema.jsonld Draft: Generates a clean WebSite or WebPage JSON-LD template populated only with verified on-page metadata.
4. generate_whitelabel_pdf_report
Renders a print-ready, 4-page A4 White-Label PDF Proposal via offline headless Chromium (templates/white_label_report.html for English, templates/white_label_report_zh.html for Simplified Chinese), complete with:
Page 1: Executive Readiness Scorecard, Measured Weight Coverage Badge, Audit Scope Metadata, and Optional Commercial Scenario Box.
Page 2: Priority findings and page-level evidence.
Page 3: AI search crawler and training crawler access matrix.
Page 4: Remediation recommendations, custom agency packages, and methodology limits.
Every generated PDF is paired with a <filename>.audit.json sidecar written without overwriting existing exports and SHA-256 hash so every number on the page can be traced back to raw HTTP evidence.
What You Get in the Download Package (geo-seo-mcp-2.1-commercial-gumroad-clean.zip)
Full Unencrypted Python Source Code (server.py & core/ modules): Audit, customize, or extend the scoring weights and HTML/CSS PDF templates for your agency workflow.
Bilingual White-Label HTML/CSS Templates (templates/): Clean A4 print layouts in both English (en) and Simplified Chinese (zh-CN).
One-Click Windows PowerShell Installer (install.ps1): Automatically provisions.venv, installs locked dependencies (requirements.txt), downloads the isolated headless Chromium shell into.browser-runtime, and runs offline verification checks.
Regression Test Suite (tests/): Unit, schema, security, and stdio MCP protocol verification scripts.
Sample Deliverables (sample_deliverables/): Real 4-page rendered A4 PDF sample report, 4 individual page preview PNGs, a 4-page contact sheet, and sample robots.txt, llms.txt, and schema.jsonld outputs.
Buyer Quick-Start Guide (README.md & README_ZH.md): Copy-paste JSON configurations for Claude Desktop and Cursor plus ready-to-run agency prompt workflows.
System Requirements & Technical Boundaries
Operating System & Runtime: Windows 10/11 with Python 3.11 or Python 3.12 (64-bit) and PowerShell. (For a manual macOS/Linux setup, create a standard venv with pip install -r requirements.txt and playwright install --only-shell chromium; platform behavior depends on your environment).
MCP Client Compatibility: Any MCP client supporting stdio transport, including Claude Desktop, Cursor, Windsurf, Cline, and custom Python/TypeScript MCP hosts.
Honest Technical Guardrails:
Audits publicly reachable HTTP/HTTPS URLs on standard ports (80/443); blocks localhost, private intranet IPs, and SSRF vectors automatically.
Respects WAF/Cloudflare 403/429 challenges without attempting illegal anti-bot bypasses; blocked items are transparently reported as unmeasured.
Headless DOM observation runs a strict 2-second window (up to 24 sub-resources / 12 MiB) to capture SPA shells safely without hanging your MCP client.
中文说明
GEO and SEO Readiness MCP Server 2.1(商业机构白标版) 是专为 SEO 代运营机构、技术营销顾问与增长工程师打造的本地 Python Model Context Protocol(stdio 传输协议)服务端。审计输出保留证据引用,PDF 导出附带审计数据伴随文件与 PDF SHA-256 校验值。
12 项商业能力
审计目标范围: 同时支持根域名全站抽样与具体商业落地页 URL(完整保留路径与 Query 参数)
SPA 动态网页与前端框架: 内置有界 2 秒无头 Chromium DOM 真实渲染观测(支持 render_mode=auto/always/never)
AI 爬虫精细化分类: 严格区分 AI 实时搜索爬虫(OAI-SearchBot、PerplexityBot、Claude-SearchBot)与 AI 训练爬虫(GPTBot、ClaudeBot、Google-Extended)
拦截与未测数据处理: 遇到 403/429、截断或 DOM 未完成时诚实标记为 unmeasured(未测),并显式输出 measured_weight(实测权重占比)
竞品对标矩阵: 对客户与竞品执行真实双站审计,校验口径与网络通道一致性后计算带符号 score_gap(客户总分 - 竞品总分)
大型 XML 站点地图: 支持 4 MiB 流式有界读取与 gzip 解压(基于 defusedxml 安全提取前 2,000 条闭合 <loc> 并防 XML 炸弹)
网络与代理通道适配: 支持 GEO_PROXY_MODE=auto/fallback/direct 三模式,原生解析 HTTP_PROXY、HTTPS_PROXY、ALL_PROXY 与 NO_PROXY
商业价值测算口径: 仅在显式传入 monthly_visitors 与 avg_deal_usd 时输出公式透明的理论上限商业价值估算,绝不伪造实际损失
robots.txt 补丁安全性: 全量保留原站规则与 Sitemap 声明,仅在末尾追加缺失的 AI 搜索爬虫放行规则;原文件截断或异常时拒绝生成覆盖版
客户交付物形态: 4 页印刷级 A4 白标 PDF 客户提案报告(内置英/中双语独立模板)+ 完整.audit.json 证据伴随文件
机构品牌与报价定制: 支持将你的 agency_name、联系方式与自定义 AgencyPackage 服务套餐报价表直接渲染进 PDF 第 4 页
工程质量与回归验证: 包含回归测试与 MCP stdio 协议检查脚本。
开箱即用的 4 大异步 MCP 核心工具
1. audit_website_geo_readiness(五维加权 GEO/SEO 就绪度审计):
基于 readiness-1.1 五维加权评分体系(搜索爬虫准入 30%、JSON-LD 结构化数据 25%、llms.txt 可读性规范 15%、XML Sitemap 完整性 15%、首字节响应速度 TTFB 15%)对主站或落地页执行真实取证,并生成带唯一 audit_id 的不可变快照。
2. compare_competitors_geo(真实双站竞品差距矩阵):
加载或实时执行客户与竞品双站审计,生成包含 5 大维度证据引用路径(<audit_id>.dimensions.<key>.score)的对标矩阵,并在双站实测维度、评分版本、审计范围与网络通道一致时给出精确总分差(score_gap)。
3. generate_client_pitch_and_patches(客户诊断话术与三大安全修复补丁):
基于真实审计快照输出按优先级排序的技术阻碍诊断、可选的透明商业价值估算,以及保存在 snapshots/patches/<audit_id>/ 下的三份待审阅补丁文件(完整保留原规则的 robots.txt、基于真实抓取引用的 llms.txt 草案、基于真实页面元数据的 schema.jsonld 草案)。
4. generate_whitelabel_pdf_report(4 页 A4 白标 PDF 提案报告生成):
调用本地离线 Chromium 渲染 4 页 A4 商业提案 PDF(支持 report_language="en" 英文版与 "zh-CN" 中文版),自动附上同名.audit.json 证据包与 SHA-256 校验值,且默认防覆盖已有文件。
### 安装与交付说明
两档均交付同一个完整源码 ZIP。Solo(49 美元)为单人授权,支持个人及自由职业客户审计;Agency(129 美元)支持机构团队使用、客户白标 PDF、品牌联系方式及套餐表。价格为一次性美元价格,适用税费另计。需要 Python 3.11/3.12、支持 stdio 的 MCP 客户端及依赖和 Chromium 安装。Windows 提供 PowerShell 安装脚本,macOS/Linux 可按手册手动配置。AI 客户端订阅、托管与代运营服务不包含在售价内。
FAQ / 常见问题
Q1: What does Commercial Version 2.1 include?
English Answer: Commercial Version 2.1 includes bounded 2-second headless Chromium DOM observation for JavaScript SPAs, full landing-page URL auditing (preserving path and query strings), HTTP/HTTPS proxy routing (GEO_PROXY_MODE), 4 MiB streaming for large.xml.gz sitemaps, non-destructive robots.txt patching, real two-site competitor benchmarking (evidence-linked score gaps), and offline 4-page White-Label A4 PDF report generation in English and Simplified Chinese.
中文对照:商业版 v2.1 包含面向 SPA 动态网页的 2 秒有界无头 Chromium DOM 观测、完整保留路径与查询参数的落地页级审计、HTTP/HTTPS 代理路由(GEO_PROXY_MODE)、支持.xml.gz 的 4 MiB 大型站点地图流式解析、不破坏原规则的 robots.txt 安全补丁、真实双站竞品对标(彻底杜绝硬编码 86 分),以及支持英/中双语的 4 页离线白标 A4 PDF 客户提案报告生成。
Q2: Which AI crawlers does the Search Crawler Access dimension evaluate?
English Answer: The Search Crawler Access dimension (30% of the total score) uses urllib.robotparser to evaluate live AI search bots (OAI-SearchBot, PerplexityBot, and Claude-SearchBot) across every sampled URL. It tracks AI training crawlers (GPTBot, ClaudeBot, and Google-Extended) separately so that websites blocking model training are never unfairly penalized on their AI search visibility score.
中文对照:搜索爬虫准入维度(占总分 30% 权重)使用 Python 标准库 urllib.robotparser 对所有抽样页面评估三大实时 AI 搜索爬虫(OAI-SearchBot、PerplexityBot 与 Claude-SearchBot)。同时,它将 AI 模型训练爬虫(GPTBot、ClaudeBot、Google-Extended)单独分离统计——网站即便在 robots.txt 中禁止大模型抓走语料用于训练,也绝不会因此被扣减 AI 搜索可见度得分。
Q3: Can this MCP server audit client-rendered React, Vue, or Next.js single-page applications (SPAs)?
English Answer: Yes. With render_mode="auto" (default) or render_mode="always", the server launches an isolated headless Chromium instance using a strict 2-second observation window (capped at 24 sub-resources and 12 MiB) to inspect the rendered DOM and extract dynamically injected <script type="application/ld+json"> schemas. If a dynamic shell fails to finish rendering within the window, the schema dimension is transparently marked unmeasured rather than scored as a false zero.
中文对照:支持。在默认的 render_mode="auto" 或强制渲染模式 render_mode="always" 下,服务端会拉起项目隔离的无头 Chromium 实例,执行严格受控的 2 秒 DOM 渲染观测(上限 24 个子资源、12 MiB 流量),精准提取由前端 JS 动态注入的 JSON-LD 结构化数据。若动态空壳在 2 秒窗口内未能完成渲染,Schema 维度会诚实标记为 unmeasured(未测),绝不误判为虚假的 0 分。
Q4: What happens if a target website blocks automated requests with a 403, 429, or CAPTCHA challenge?
English Answer: The server strictly respects website access controls and never attempts illegal anti-bot or CAPTCHA bypasses. Any dimension that cannot be verified due to HTTP 403/429 status codes, response truncation, or incomplete DOM rendering is recorded with score=None and availability="unmeasured". The report prominently displays the exact measured_weight percentage so your agency never presents partial data as a full site audit.
中文对照:本工具严格遵守目标网站的访问控制策略,绝不执行非法的反爬绕过或验证码破解。任何因 HTTP 403/429 状态码、响应体超限截断或 DOM 未完成而无法核实的维度,都会被记录为 score=None 与 availability="unmeasured"(未测)。生成的报告会醒目展示本次审计的实际有效权重占比(measured_weight),确保代运营机构绝不会把残缺数据当成完整结论交付给客户。
Q5: Will the generated robots.txt patch overwrite or break my client's existing SEO rules?
English Answer: It generates a candidate file for review; it does not deploy changes to your website. The generate_client_pitch_and_patches tool reads the client's existing live robots.txt in full, preserves every current User-agent, Allow, Disallow, and Sitemap directive verbatim, and appends only the missing AI search crawler rules at the end of the candidate file. If the original robots.txt was truncated (>4 MiB) or unparseable, automatic patch generation is blocked and flagged for manual review.
中文对照:工具输出待审阅的候选文件,不会直接部署到你的网站。generate_client_pitch_and_patches 工具会完整读取客户现有的线上 robots.txt,原封不动地保留原有的每一条 User-agent、Allow、Disallow 与 Sitemap 声明,仅在候选文件末尾追加缺失的 AI 搜索爬虫放行规则。如果原站的 robots.txt 超出 4 MiB 被截断或格式损坏,系统会自动熔断并禁止生成覆盖文件,提示人工介入审阅。
Q6: How do I customize the 4-Page White-Label PDF Report with my agency's branding and pricing?
English Answer: Pass your agency_name, agency_contact, and a list of AgencyPackage objects (name, price, timeline, and deliverables) when calling generate_whitelabel_pdf_report. Set report_language="en" for the English A4 template or "zh-CN" for the Simplified Chinese A4 template. The server renders the PDF locally via offline Chromium without sending client data to any third-party cloud PDF API.
中文对照:在调用 generate_whitelabel_pdf_report 时,直接传入你的机构名称(agency_name)、联系方式(agency_contact)以及自定义的 AgencyPackage 服务套餐数组(包含套餐名 name、价格 price、周期 timeline 与交付项列表 deliverables)即可。设置 report_language="en" 即输出全英文 A4 提案报告,设置 "zh-CN" 即输出简体中文 A4 提案报告。整个 PDF 渲染由本地离线 Chromium 完成,绝不向任何第三方云端 PDF 接口上传客户敏感数据。
Q7: Does the server support HTTP/HTTPS proxies for auditing geo-restricted or cross-border websites?
English Answer: Yes. Set GEO_PROXY_MODE in your MCP environment configuration to auto (uses HTTPS_PROXY, HTTP_PROXY, or ALL_PROXY automatically while respecting NO_PROXY), fallback (connects directly first and retries via proxy on failure), or direct (disables proxy routing). Both the HTTP fetcher and the headless Chromium resource loader share the exact same proxy and SSRF security checks.
中文对照:完全支持。只需在 MCP 启动环境变量中将 GEO_PROXY_MODE 设为 auto(优先读取环境中的 HTTPS_PROXY / HTTP_PROXY / ALL_PROXY 并遵守 NO_PROXY)、fallback(先直连,失败后自动切换代理重试)或 direct(强制直连)。底层 HTTP 抓取器与无头 Chromium 浏览器资源加载器共享完全一致的代理通道与公网安全检查,且绝不在报告或日志中泄露代理账号密码。
Q8: Do I get full source code access, and can I modify the scoring weights or PDF styling?
English Answer: Yes. Both the Solo Consultant Edition ($49) and the Agency White-Label Edition ($129) include 100% unencrypted Python source code (server.py, core/*.py) and editable Jinja2 HTML/CSS print templates (templates/white_label_report.html and templates/white_label_report_zh.html). You can adjust CSS colors, fonts, scoring weights, or pitch templates freely.
中文对照:是的。无论购买独立顾问版($49)还是机构白标商业版($129),均交付 100% 未加密的完整 Python 源代码(server.py、core/*.py)以及可直接编辑的 Jinja2 HTML/CSS 印刷级模板(templates/white_label_report.html 与 templates/white_label_report_zh.html)。你可以自由修改报告配色、字体、评分权重或商业话术逻辑。
Methodology and Deployment Boundaries
The readiness-1.1 weights are an engineering heuristic, not a search-engine ranking formula. Scores do not guarantee indexing, AI citations, traffic, rankings, revenue, or client contracts. Optional commercial scenario math uses your supplied inputs and is not a measured revenue loss. Audit access is subject to the target website's controls. Review every generated patch before deployment. Audit sidecars provide traceability; they are not cryptographically signed proof against modification.
方法边界
readiness-1.1 权重是工程启发式模型,不是搜索引擎排名公式。不承诺收录、AI 引用、流量、排名、营收或签单。可选商业测算依赖买家输入,不能视作实际营收损失。补丁部署前必须人工审阅;伴随数据用于追溯,不等于数字签名或防篡改认证。
NEW — Hosted Upgrade Pack included (2026-10-27): This listing now also includes geo-seo-mcp-hosted-20261027.zip, adding Hosted Storefront / Target Landing Page scoring (hosted-landing-1.0): score hosted product pages and landing pages with a dedicated rubric (robots 30%, visible semantics 30%, product schema 25%, observed TTFB 15%). Also included: stabilized DOM observation (layout-visible text, up to 45s), full H1–H3 and JSON-LD retention, and source/rendered schema deduplication. Site scoring remains readiness-1.1; hosted and site scores are not comparable. 91 automated tests pass. The original 2.1 clean package remains included unchanged.