Screpy SEO MCP

Connect a compatible AI client to Screpy’s hosted Model Context Protocol server with browser-based OAuth—no local server or REST API key required. Use 44 structured tools to investigate project-scoped crawl evidence, pages, links, images, Quick Wins, Rank Tracker, stored AI Visibility, Core Web Vitals, and uptime, with explicit intent and confirmation protecting high-impact actions.

TL;DR

Too Long; Didn't Read

What Is an SEO MCP Server?

Screpy SEO MCP is a hosted remote MCP server that exposes 44 structured Screpy tools to compatible AI clients. Add https://mcp.screpy.com, complete browser-based OAuth, and the client can work only with projects available to that member. Read tools support investigation; supported write tools require clear user intent, and capacity-consuming or destructive actions require confirmation.

Connection and project access

• Add the hosted remote server URL https://mcp.screpy.com to a client that supports remote HTTP MCP servers and browser-based OAuth.
• Screpy provides setup guides for ChatGPT, Claude, Codex, Cursor, VS Code with GitHub Copilot, and Gemini CLI.
• MCP uses its own OAuth connection; a Screpy REST API key cannot be used for the MCP sign-in flow.
• Every tool call is scoped to projects owned by or shared with the authenticated member, and project access is checked again when the tool runs.
• The connection can be revoked from Connected Apps, after which the client must be authorized again before it can use Screpy tools.

Read and investigate

• Account context and plan usage show available project access and capacity before an assistant starts a larger workflow.
• Project, crawl, on-page, and comparison tools move from website-level context to one completed crawl and one exact URL.
• Page, link, and image tools support structured filters and source-page tracing so findings remain tied to crawl evidence.
• AI Visibility tools read stored overview, mention, answer, citation, prompt, and source evidence without starting a new analysis.
• Quick Wins, Core Web Vitals, and uptime tools add existing opportunity, field-performance, and availability context without mixing those sources into crawl records.

Manage supported workflows safely

• Rank Tracker tools read options, overview metrics, keywords, stored analyses, results, segments, previews, and unmatched-keyword context.
• AI Visibility tools add prompts only after explicit intent and delete a project prompt only after confirmation.
• Keyword additions, keyword deletion, and destructive segment actions require explicit user intent and the required confirmation before execution.
• Project creation, settings changes, project deletion, and starting a crawl are available only through their dedicated tools and controlled action rules.
• Available data and actions still depend on the member’s plan, remaining allowance, project state, feature access, and connected sources.
• MCP does not provide Rank Tracker exports, live SERP lookup, or manual reanalysis of an existing tracked keyword.
• MCP does not provide AI Visibility CSV or Excel exports, prompt suggestions, or manual prompt reanalysis.

Connect the hosted SEO MCP server in four steps.

Screpy keeps setup short: add one remote endpoint, authorize the correct account, and verify project access before asking for deeper analysis or an approved change.

1. 01

   Choose a compatible AI client

   Use one of Screpy’s current setup guides or another client that supports remote HTTP MCP servers and browser-based OAuth.
2. 02

   Add the hosted Screpy endpoint

   Enter https://mcp.screpy.com in the client. You do not need to install a local server or copy a Screpy REST API key.
3. 03

   Complete browser-based OAuth

   Sign in with the Screpy account that can access the projects you want to investigate, then approve the connection.
4. 04

   Verify the project scope

   Ask the client to “List my Screpy projects” before a larger workflow so you can confirm the account and available project access.

Connect the AI client you already use.

Screpy provides current setup guides for these six clients. Another client may also work when it supports remote HTTP MCP servers and browser-based OAuth, but verify its current transport and authorization support before connecting.

ChatGPT

Add Screpy as a custom MCP app, complete browser-based OAuth, and use project-scoped tools inside ChatGPT conversations.

Codex

Bring crawl, URL, Rank Tracker, and monitoring context into Codex before planning or implementing a code-level SEO change.

Claude

Connect Screpy through Claude’s remote connector support and investigate completed crawls, exact URLs, and project health.

Gemini CLI

Authorize Screpy through OAuth in Gemini CLI and use authenticated SEO tools from a terminal-based agent workflow.

Cursor

Add Screpy as a remote HTTP MCP server and keep crawl evidence beside the code that controls the affected website.

VS Code and GitHub Copilot

Connect Screpy through VS Code MCP support and use project data from GitHub Copilot Chat while reviewing implementation work.

Connect once, then keep every answer tied to the right Screpy account.

Add the hosted remote endpoint, complete browser-based OAuth, and verify access by listing projects. The AI client discovers structured Screpy tools without a local package, copied REST key, or manual dashboard export.

• Use https://mcp.screpy.com as the remote server URL
• Authorize the intended Screpy account in the browser
• List accessible projects before selecting a project UID
• Revoke the client later from Connected Apps when needed

Create a focused review of hosted mcp connection.

Add the hosted remote endpoint, complete browser-based OAuth, and verify access by listing projects. The AI client discovers structured Screpy tools without a local package, copied REST key, or manual dashboard export.

Move from website health to the exact crawl record behind a finding.

Start with the selected project and completed crawl, then narrow the investigation through crawl summaries, on-page metrics, structured page filters, exact URL details, and source relationships.

• List completed crawls and check whether a current crawl is ready to investigate
• Review crawl-level health indicators, issue counts, and changes between completed crawls
• Filter the page inventory, then retrieve detailed data for one exact crawl URL
• Trace link and image findings to the pages where they appear

Create a focused review of crawl investigation.

Start with the selected project and completed crawl, then narrow the investigation through crawl summaries, on-page metrics, structured page filters, exact URL details, and source relationships.

Investigate ranking movement before changing tracked keywords or segments.

Read visibility, position distribution, keyword details, stored history and results, and segment matches. Preview a grouping first; save, add, or delete only when the user explicitly asks for that exact action.

• Read overview metrics and current plan usage
• Filter keyword records and open one keyword’s stored history
• Preview segment rules and overlap before saving them
• Require explicit intent for capacity-consuming or destructive changes

Create a focused review of rank tracker workflow.

Read visibility, position distribution, keyword details, stored history and results, and segment matches. Preview a grouping first; save, add, or delete only when the user explicitly asks for that exact action.

Trace stored AI visibility from overview to source evidence.

Read visibility trends, mentions, answers, citations, prompts, and source activity. Add prompts only with explicit user intent and require confirmation before deleting one.

• Read stored overview and answer evidence
• Follow prompt, citation, domain, and page activity
• Use opaque cursors for longer result sets
• Keep exports, suggestions, and manual reanalysis outside MCP

Create a focused review of ai visibility workflow.

Read visibility trends, mentions, answers, citations, prompts, and source activity. Add prompts only with explicit user intent and require confirmation before deleting one.

Add existing opportunities, field performance, and availability to the investigation.

Quick Wins, Core Web Vitals, and uptime tools provide separate project-scoped evidence. Use them beside crawl and page findings while preserving the meaning and availability limits of each source.

• Retrieve existing Search Console Quick Wins by category and range
• Review available Core Web Vitals results separately from crawl data
• Check current uptime status and recent incidents
• Treat missing monitoring data as unavailable, not as a healthy result

Create a focused review of opportunities and health.

Quick Wins, Core Web Vitals, and uptime tools provide separate project-scoped evidence. Use them beside crawl and page findings while preserving the meaning and availability limits of each source.

44 project-scoped tools for investigation and controlled action.

The current member MCP server groups 44 tools into account, project, crawl, on-page, opportunity, health, Rank Tracker, and AI Visibility workflows. Availability still depends on account access, project state, connected data, and plan allowance.

Account context and usage

Read the authenticated member’s product access, available project count, current usage, and remaining allowance before a workflow consumes capacity.

Project discovery and overview

List accessible projects, open one project’s high-level SEO context, or compare summary metrics across two to five available projects.

Controlled project management

Create a project, update allowed settings, or delete a selected project only through the dedicated tool and the required user-intent or confirmation flow.

Crawl selection and status

List crawl runs for one project, select the exact crawl UID, and check progress or state before treating results as complete.

Controlled crawl start

Start a new crawl only for the selected project and only after explicit confirmation because the action may consume plan allowance.

Crawl summary and comparison

Read crawl totals, health indicators, issue counts, and on-page metrics, then compare two completed crawls to inspect metric changes.

Page inventory and exact URL

Filter pages within one crawl and retrieve detailed crawl data for one exact URL after the relevant record has been identified.

Links and source pages

List link records with structured filters and trace one destination back to the crawl pages that contain it.

Images and source pages

List image records from one crawl and return the pages that reference a selected image for contextual investigation.

Search Console Quick Wins

Retrieve existing project Quick Wins by supported category, reporting range, and cursor without presenting them as a full Search Console replacement.

Core Web Vitals and uptime

Read available field-performance results, current uptime status, and recent incidents as separate health signals for the selected project.

Rank Tracker setup and overview

Read supported countries, languages, devices, frequencies, segment-rule options, overview metrics, visibility, distribution, and plan usage.

Keyword inventory and detail

Filter and paginate tracked keywords, then open one keyword with up to 30 stored analyses and the results from the selected or latest analysis.

Controlled keyword changes

Add or delete up to 100 project-scoped keyword IDs per action only with explicit confirmation, including the requested device variants.

Segment list and preview

List segments with SQL-backed match counts and unmatched keywords, then preview proposed rules, overlap, average position, and example URLs without saving.

Controlled segment management

僅在使用者明確要求儲存或移除確切規則時，才建立、更新或刪除專案範圍的區隔。

AI 能見度證據

讀取已儲存的概覽、提及、回答、引用、提示、來源與活動記錄，不觸發新的分析。

受控的 AI 提示變更

僅在明確的使用者意圖下，為單一市場新增 1 到 20 個提示，並在軟刪除專案提示前要求確認。

提出有界限的問題，以保留證據。

一個強大的 MCP 提示會指明專案、來源、範圍與動作邊界。以下範例明確指出所請求的證據以及不得執行的動作。

調查發布回歸

「列出我的專案，選擇 [site]，比較最近兩次完成的爬蟲，並顯示變更的指標。不要開始新的爬蟲。」

• 現有的爬蟲證據
• 可比較的爬蟲摘要
• 不消耗容量的動作

檢視單一確切 URL

「開啟 [site] 最近完成的爬蟲，找到 [URL]，解釋其頁面、連結與圖片證據，並區分事實與假設。」

• 確切 URL 查詢
• 來源頁面脈絡
• 先證據後建議

解釋排名變動

「顯示 [site] 的能見度、下降的關鍵字與受影響的排名 URL。不要新增、刪除或重新分析關鍵字。」

• 投資組合能見度
• 已儲存的關鍵字分析
• 唯讀調查

預覽實用的區隔

「預覽 [rules] 的 Rank Tracker 區隔，顯示與現有區隔的相符與重疊，且不要儲存它。」

• 規則預覽
• 相符與不相符的關鍵字
• 不儲存變更

調查 AI 能見度

「摘要 [site] 已儲存的 AI 能見度，然後顯示最大變更背後的提示、回答、引用與來源。不要新增或刪除提示。」

• 已儲存的回答證據
• 提示與來源向下鑽取
• 唯讀調查

將存取與高影響動作保留在使用者控制之下。

MCP 讓工具可供 AI 用戶端使用；它不會消除驗證所選帳戶、檢視工具呼叫、保護容量與確認變更的需求。

OAuth 與專案範圍存取

連線使用瀏覽器型 OAuth，且每次工具呼叫都會檢查所選專案的存取權，而非暴露無關的 Screpy 資料。

• 不複製 REST API 金鑰
• 每次呼叫檢查成員存取權
• 可從已連線應用程式撤銷

用於調查的讀取工具

專案、爬蟲、頁面、連結、圖片、Quick Wins、Rank Tracker、AI 能見度、Core Web Vitals 與正常運行時間工具可以收集證據，而不變更專案狀態。

• 選擇明確的專案
• 必要時使用已完成的資料
• 保持來源限制可見

支援寫入的明確意圖

專案、爬蟲、關鍵字、區隔與 AI 能見度提示變更使用專用工具。請求必須清楚描述變更，而非將其隱藏在廣泛的分析提示中。

• 明確的目標與動作
• 專用管理工具
• 計畫與容量檢查

高影響動作的確認

消耗容量或破壞性的動作需要確認。已連線的 AI 用戶端也應讓您檢查並拒絕工具呼叫。

• 執行前確認
• 檢視工具請求
• 無 AI 能見度匯出或手動重新分析

為工作選擇 MCP、REST API 或儀表板。

使用 MCP 進行對話式、工具型調查；使用 REST API 進行確定性的產品整合；使用儀表板進行直接視覺檢閱與手動控制。

Screpy SEO MCP

透過託管的遠端 MCP 伺服器，將相容的 AI 用戶端連線至已驗證、專案範圍的 Screpy 工具。

連線 MCP

Screpy REST API

使用 API 憑證、已記錄的端點、分頁與速率限制，建立明確的應用程式與自動化流程。

探索 SEO API

Screpy 儀表板

直接在產品介面中檢閱專案、爬蟲、頁面、排名、機會、效能與監控。

開啟 Screpy

Screpy 文件

安全地連線、授權並使用 Screpy MCP

遵循 MCP 指南以連線相容的 AI 用戶端、完成瀏覽器型 OAuth、驗證專案存取、撰寫有界限的提示，並疑難排解連線。

• 檢閱 MCP 概覽
• 了解 OAuth 與存取
• 使用範例 SEO 提示
• 疑難排解連線

遵循 MCP 設定指南

SEO MCP 伺服器常見問題

了解遠端設定、OAuth 存取、支援的用戶端、工具範圍、確認、Rank Tracker 限制、資料隱私、撤銷，以及 MCP 與 REST API 之間的差異。

什麼是 SEO MCP 伺服器？

SEO MCP 伺服器透過 Model Context Protocol 向相容的 AI 用戶端暴露結構化的 SEO 工具。Screpy SEO MCP 將已驗證的成員帳戶連線至專案範圍的工具，用於網站爬蟲、頁面記錄、連結、圖片、Quick Wins、Rank Tracker、Core Web Vitals、正常運行時間與受控的專案動作。

如何連線至 Screpy SEO MCP？

選擇支援遠端 HTTP MCP 伺服器與瀏覽器型 OAuth 的用戶端，新增 https://mcp.screpy.com, 完成 Screpy 登入與核准流程，然後要求用戶端列出您的 Screpy 專案。看到預期的專案即確認連線已就緒。

我需要安裝本機 MCP 伺服器或建立 API 金鑰嗎？

不需要。Screpy 提供託管的遠端 MCP 端點，因此支援的用戶端可以在不執行本機套件的情況下連線。MCP 使用瀏覽器型 OAuth；Screpy REST API 金鑰是獨立的憑證，不能用於 MCP 連線。

哪些 AI 用戶端有 Screpy 設定指南？

Screpy 目前記錄了 ChatGPT、Claude、Codex、Cursor、VS Code 搭配 GitHub Copilot 與 Gemini CLI 的設定。其他用戶端在支援相容的遠端 HTTP 傳輸與 OAuth 時可能可以運作，但其連線流程應根據用戶端目前的 MCP 實作進行驗證。

Screpy SEO MCP 提供多少工具？

目前的成員 MCP 伺服器註冊了 44 個工具，涵蓋帳戶脈絡、專案、爬蟲、頁面記錄、Quick Wins、網站健康、Rank Tracker 與 AI 能見度。結果中可用的工具仍取決於成員的存取權、專案狀態、已連線的資料、方案與剩餘額度。

Screpy MCP 如何保護專案存取？

用戶端透過 OAuth 連線，且僅接收由已驗證成員擁有或與其共用的專案。Screpy 在工具執行時再次檢查專案存取權，因此成員失去存取權後，專案無法繼續使用。

Screpy MCP 工具是唯讀的嗎？

大多數調查工具讀取現有資料，但伺服器也包含專案、爬蟲、Rank Tracker 關鍵字與區隔的受控動作。寫入工具應僅在使用者明確要求該確切變更時執行。消耗容量與破壞性的動作需要確認。

哪些 MCP 動作需要確認？

開始爬蟲、建立或刪除專案、新增或刪除 Rank Tracker 關鍵字，以及破壞性的區隔動作，會使用適合該作業的明確意圖或確認控制。用戶端應顯示建議的工具呼叫，以便使用者在執行前核准或拒絕。

AI 助理可以調查單一確切 URL 嗎？

可以。助理可以選擇已完成的爬蟲、篩選其頁面庫存，並請求單一確切 URL 的詳細爬蟲資料。個別工具可以列出連結與圖片，並將選定的目的地或圖片追溯至其來源頁面。

Screpy MCP 可以比較網站爬蟲或專案嗎？

可以。一個工具比較兩到五個可存取專案的摘要指標，而另一個工具比較兩個選定的已完成爬蟲。使用專案比較取得網站層級的脈絡，使用爬蟲比較取得技術 SEO 快照之間的變更。

Screpy MCP 提供完整的即時 Search Console 報告嗎？

不。MCP 工具會為支援的類別與報告範圍傳回現有的 Search Console Quick Wins。它不會重現每個 Search Console 儀表板，也不會暴露獨立的無限制即時報告介面。

Screpy MCP 可以對 Rank Tracker 做什麼？

它可以讀取支援的選項、概覽指標、能見度、位置分佈、關鍵字清單、單一關鍵字的已儲存分析與結果、區隔清單與區隔預覽。它也可以在明確請求時新增或刪除關鍵字，並建立、更新或刪除區隔。

哪些 Rank Tracker 動作無法透過 MCP 使用？

MCP 不會匯出 Rank Tracker 資料、執行即時 SERP 查詢，或手動重新分析現有的追蹤關鍵字。新關鍵字開始其正常首次分析，現有關鍵字則繼續依設定的排程執行。當工作流程需要 MCP 工具集之外的功能時，請使用 REST API 或儀表板。

Screpy MCP 可以對 AI 能見度做什麼？

它可以讀取已儲存的選項、概覽指標、提及、回答、引用、提示與來源活動。它也可以在明確的使用者意圖後新增提示，並在確認後刪除提示。它不會匯出 CSV 或 Excel 檔案、建議提示，或手動重新分析提示。

MCP 可以在調查中包含 Core Web Vitals 與正常運行時間嗎？

可以。個別工具會為所選專案傳回可用的 Core Web Vitals 結果、目前正常運行時間狀態與近期事件。如果監控資料不可用，助理應報告該狀態，而非推斷網站快速或可用。

傳回給我的 AI 用戶端的資料會發生什麼事？

Screpy 會傳回授權用戶端工具呼叫所請求的專案資料。已連線的 AI 提供者或用戶端可能根據其自身的產品設定與條款處理或保留該結果。在連線敏感專案前，請檢閱用戶端的隱私與資料控制選項。

如何撤銷 AI 用戶端的 Screpy 存取權？

在 Screpy 中開啟已連線應用程式，找到已授權的 AI 用戶端，並撤銷連線。用戶端在完成新的瀏覽器型 OAuth 核准前，無法再次使用 Screpy MCP。

Screpy MCP 與 Screpy REST API 有何不同？

MCP 專為相容的 AI 用戶端設計，可在對話或編碼工作流程中探索並呼叫較小、以代理為焦點的工具集。REST API 專為需要直接 HTTP 請求、API 金鑰、分頁與程式化控制的自訂產品、儀表板、指令碼與自動化設計。