Docs · MCP server

CitedRank MCP server

The CitedRank MCP server is a free, remote Model Context Protocol server that gives Claude Code, Codex, Cursor and other AI agents 9 tools: SEO and GEO audits, sitemap discovery, page scraping, and Sitemap Monitor — which tells your agent what new pages the competitor sites you watch have published.

Endpoint
https://citedrank.co/mcp
Transport
Streamable HTTP (JSON responses)
Auth
Bearer personal API key (crk_…)
Protocol versions
2025-06-18 · 2025-03-26 · 2024-11-05
Tools
9 (3 for Sitemap Monitor)
Price
Free with Google sign-in

1. Create a personal API key

Your key identifies you to the MCP server — monitor tools return only the sites you watch. Create it once, paste it into your client below, and keep it private like a password.

2. Add CitedRank to your client

Pick your client. Each snippet uses crk_your_api_key until you create a key above — then your key is filled in everywhere on this page.

Claude Code

Run once in your terminal (--scope user makes it available in every project):

claude mcp add --transport http --scope user citedrank https://citedrank.co/mcp \
  --header "Authorization: Bearer crk_your_api_key"

Check the connection:

claude mcp list

Codex

Add to ~/.codex/config.toml:

[mcp_servers.citedrank]
url = "https://citedrank.co/mcp"
bearer_token_env_var = "CITEDRANK_API_KEY"

Put the key in your shell profile (~/.zshrc or ~/.bashrc), open a new terminal, then start Codex:

export CITEDRANK_API_KEY="crk_your_api_key"

Cursor

Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "citedrank": {
      "url": "https://citedrank.co/mcp",
      "headers": {
        "Authorization": "Bearer crk_your_api_key"
      }
    }
  }
}

VS Code

Add to .vscode/mcp.json in your workspace:

{
  "servers": {
    "citedrank": {
      "type": "http",
      "url": "https://citedrank.co/mcp",
      "headers": {
        "Authorization": "Bearer crk_your_api_key"
      }
    }
  }
}

Claude Desktop

Claude Desktop runs local (stdio) servers, so bridge to ours with mcp-remote (needs Node.js). Add to claude_desktop_config.json, then restart Claude Desktop:

{
  "mcpServers": {
    "citedrank": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://citedrank.co/mcp",
        "--header",
        "Authorization:${CITEDRANK_AUTH}"
      ],
      "env": {
        "CITEDRANK_AUTH": "Bearer crk_your_api_key"
      }
    }
  }
}

Any other MCP client

Add a remote server using the Streamable HTTP transport with these two values:

URL:    https://citedrank.co/mcp
Header: Authorization: Bearer crk_your_api_key

3. Ask your agent

You don’t call tools by name — describe what you want and the agent picks the right one. For example:

  • “What new pages did my competitors publish in the last 7 days?”
  • “Start watching competitor.com/blog/ and check it every 6 hours.”
  • “List the new pages on competitor.com this month and run a GEO audit on the most interesting one.”
  • “Run an SEO audit on https://example.com/pricing and tell me the three biggest fixes.”
  • “Get every URL in stripe.com's sitemap and group them by section.”
  • “Scrape the top 10 posts from news.ycombinator.com as JSON.”

Tools reference

The same list your client receives from tools/list.

Sitemap Monitor

Per-user tools — they read and change the sites watched by the key’s owner. Manage the same monitors on the Sitemap Monitor page.

monitor_new_pages

What new pages did the sites I watch publish?

List the NEW pages that appeared on the competitor sites this user watches with CitedRank Sitemap Monitor, newest first, with page titles and the time each was first seen. Only pages published after monitoring started are included (the first check is a baseline). Use when the user asks what their competitors published recently, what's new on a watched site, or for a digest of new pages. Pass site to narrow to one watched domain. Follow up with the seo or geo tools on interesting URLs.

ArgumentTypeDescription
daysnumberLook back this many days (1-365). Default 7.
sitestringOptional: only this watched site, as a domain or URL (e.g. 'competitor.com').
limitnumberMax pages to return (1-2000). Default 300.

Returns: { days, site, pages: [{ url, title, first_seen_at, lastmod, domain, monitor_id }] }

monitor_list

Which sites am I monitoring?

List the sites this user watches with CitedRank Sitemap Monitor: domain and path, how many URLs are tracked, new pages in the last 7 days, check interval, last/next check time, status and any error, and whether email alerts are on. Use to answer 'what am I monitoring' or before calling monitor_new_pages for a specific site.

No arguments.

Returns: { monitors: [{ id, target, domain, path_prefix, url_count, new_7d, interval_hours, last_checked_at, next_check_at, last_status, last_error, email_to }], limit }

monitor_add

Start watching a competitor for new pages

Start watching a website for new pages with CitedRank Sitemap Monitor. Give a domain ('competitor.com') to watch its whole sitemap, or a path ('competitor.com/blog/') to watch only that section; sites without a sitemap fall back to the links on that page. The first check (a baseline) runs immediately; new pages are reported from the next check on and emailed to the user. Free plan: up to 20 sites, at most 3 additions per minute.

ArgumentTypeDescription
targetrequiredstringDomain or URL to watch, e.g. 'competitor.com' or 'competitor.com/blog/'.
interval_hours6 | 12 | 24 | 72 | 168How often to check, in hours. Default 24.
email_alertsbooleanEmail the user when new pages appear. Default true.

Returns: the created monitor (same shape as an item of monitor_list).

Page & site analysis

Work on any public URL or domain.

sitemap

List every URL on a website

Discover every URL on a domain. Parses robots.txt, follows nested sitemap.xml indexes up to 4 levels deep, decompresses .gz, and probes 15+ fallback paths (WordPress, Shopify, Webflow). Returns the URL list with lastmod and priority. Use when the user asks 'what pages does X have' or before bulk-crawling a site.

ArgumentTypeDescription
domainrequiredstringDomain or URL to scan (e.g. 'stripe.com'). Only the hostname is used.

Returns: { domain, sitemaps[], total_urls, urls: [{ url, lastmod, priority }] }

crawl

Get the text content of a page

Fetch a single page and return its content as clean Markdown plus a list of image URLs. Renders JavaScript via headless Chromium. Use when the user wants the actual text content of a page (for reading, RAG ingestion, summarization). Cached: re-running on the same URL returns instantly with from_cache=true.

ArgumentTypeDescription
urlrequiredstringFull URL to fetch (https://...).
refreshbooleanForce re-render instead of using cache. Default false.

Returns: { url, status, title, word_count, html_size, mhtml_saved, from_cache, ok }

seo

Audit a URL for SEO problems

Run a 10-section SEO audit on a single URL: title length, description, canonical, headings (h1-h6), Open Graph + Twitter cards, hreflang, link graph (internal/external/nofollow), JSON-LD schema types, media (img/video/iframe + alt-text coverage), content metrics (word count, text-to-HTML ratio), tech (doctype, AMP, scripts), robots meta. Use for SEO diagnosis and competitor audits.

ArgumentTypeDescription
urlrequiredstringFull URL to audit.
refreshbooleanForce re-fetch. Default false.

Returns: a 10-section report: basics, headings, social, i18n, links, schema, media, content, tech, robots_meta.

geo

Check if a page is AI-search (GEO) ready

Run a GEO (Generative Engine Optimization) audit on a single URL — scores how easily AI search engines (ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews) can extract and cite the page. Grades seven content-quality dimensions (entity clarity, citable facts, FAQ coverage, structure, authority, comparison positioning, freshness), adjusted for the detected page type. Returns a normalized 0-100 score, a grade, the per-dimension breakdown, a core-page minimum-requirements checklist, and a prioritized fix list. Use when the user asks whether a page is ready for AI search, ChatGPT, or Google AI Overviews.

ArgumentTypeDescription
urlrequiredstringFull URL to audit for GEO.
refreshbooleanForce re-fetch. Default false.

Returns: { url, domain, page_type, score, grade, signals[] (7 dimensions, 0-10 each), requirements, recommendations[], llms_txt }

clone

Extract a site's web UI — colors, fonts, CSS

Extract a website's web UI and design system: colors (clustered by hue, primary/accent/neutral roles assigned), font families, font-size scale, border-radius scale, shadow tiers. Returns design tokens ready for export to Tailwind config / CSS variables / JSON. Use for design reverse-engineering, brand audits, building style guides. Shown in the CitedRank UI as the 'Web UI' extractor.

ArgumentTypeDescription
urlrequiredstringFull URL to extract the UI from.
refreshbooleanForce re-fetch. Default false.

Returns: { url, screenshot_available, mhtml_available, css_count, tokens, spec } — spec holds clustered design tokens (colors, typography, spacing, radii, shadows).

extract

Scrape structured rows / fields from a page

Extract structured fields from a page by applying CSS selectors. Returns a JSON array of records — one per matched item. Use when the user wants tabular data (product cards, search results, Reddit posts, GitHub repos) instead of whole-page Markdown. Three modes: 'http' (fast, no browser), 'browser' (Playwright for JS sites), 'stealth' (anti-bot evasion for Cloudflare-protected sites). Shown in the CitedRank UI as the 'Data Scraper'.

ArgumentTypeDescription
urlrequiredstringFull URL to scrape.
fieldsrequiredarrayField definitions to extract per item.
item_selectorstringCSS for repeating items (e.g. 'tr.athing'). Omit for single-record whole-page mode.
modehttp | browser | stealthFetch mode. Default 'http' (fastest).
limitnumberCap items returned (list mode only).

Returns: { url, status, item_count, duration_ms, mode_used, items: [...] }

Limits & errors

  • Sitemap Monitor, free plan: up to 20 watched sites per account, at most 3 added per minute.
  • Checks run every 6, 12, 24, 72 or 168 hours per site. The first check is a baseline; only pages that appear afterwards are “new”.
  • Up to 5 active API keys per account. Revoking a key takes effect within a minute.
  • Watched sites and webhooks must be public addresses — private and internal IPs are refused.

When a tool fails, the server returns the reason as the tool’s result (with isError: true) so your agent can read it and adjust:

MessageWhenWhat to do
Missing or invalid CitedRank API keyNo Authorization header, a typo, or a revoked key.Create a key above and send it as Authorization: Bearer <key>.
HTTP 403A free-plan cap was hit — e.g. you already watch 20 sites.Stop watching a site on the Sitemap Monitor page, then retry.
HTTP 409monitor_add on a URL you already watch.Nothing to do — it's already being monitored.
HTTP 422A bad argument: not a website address, a private IP, an unsupported interval.Use a public domain or URL; interval_hours is one of 6, 12, 24, 72, 168.
HTTP 429More than 3 sites added within a minute.Wait for the seconds given in the message, then add the next one.

Prefer scripts or an agent skill over MCP? The same tools are documented for plain HTTP in the CitedRank SKILL.md (bearer key on https://api.citedrank.co).

FAQ

What is the CitedRank MCP server?+

The CitedRank MCP server is a free, remote Model Context Protocol server at https://citedrank.co/mcp. It gives AI agents such as Claude Code, Codex, Cursor and Claude Desktop 9 tools: SEO and GEO page audits, sitemap discovery, page-to-Markdown crawling, web-UI extraction, structured scraping, and Sitemap Monitor tools that report the new pages published on competitor sites you watch.

Is the CitedRank MCP server free?+

Yes. Sign in with Google, create a personal API key and connect. The free plan includes every tool; Sitemap Monitor watches up to 20 sites per account, adding at most 3 sites a minute.

Which MCP clients does it work with?+

Any client that speaks the Streamable HTTP transport with a custom Authorization header: Claude Code, Codex, Cursor, VS Code and others. Claude Desktop connects through the mcp-remote bridge. Supported protocol versions are 2025-06-18, 2025-03-26 and 2024-11-05.

How does my agent know which competitor sites I watch?+

Your API key identifies you. The monitor tools only ever return the sites watched by the account that created the key, so the same question gives each user their own answer. A key cannot read or change anyone else's monitors.

Does it support OAuth sign-in?+

Not yet. Authentication is a personal bearer key that you create on this page and paste into your client's config. You can hold up to 5 keys and revoke any of them; a revoked key stops working within a minute.

Can my agent add sites to watch?+

Yes. monitor_add starts watching a domain or a path such as competitor.com/blog/. The first check records a baseline, and every page that appears after it is reported by monitor_new_pages and, if email alerts are on, emailed to you.