io.github.avansaber/seo-monster

Score whether a page is structured to be extracted and cited by LLM answer engines (ChatGPT, Perplexity, Claude, Google AI Overviews). Leads with a render-blindness check (AI crawlers do not run JS, so a client-rendered SPA is invisible to them), then scores evidence-backed signals: statistics, quotations, cited sources, no keyword stuffing. Reports schema.org / FAQ as informational only -- the 2026 evidence does NOT support them as AI-citation drivers, so they are not scored. Read-only HTTP GET; does not guarantee a citation.

Track brand mention + citation share-of-voice across AI answer engines (Perplexity/OpenAI/Anthropic/Gemini via their APIs, Google AI Overviews via DataForSEO) for a managed prompt set, vs competitors. Samples each prompt N times (default 7) and reports visibility with a 95% confidence interval, share-of-voice, and run-to-run volatility -- NOT an 'AI rank' (single runs are statistically meaningless). Honest bound: developer-API output differs from the logged-in consumer UI, and AIO has no API. Paid + non-deterministic; results are directional and dated, not guaranteed.

First-party view of AI traffic: (1) referral sessions from AI apps (ChatGPT, Perplexity, Gemini, Copilot, Claude, ...) via GA4, detected by both the native ai-assistant channel and a source-host regex; and (2) AI-crawler coverage -- whether robots.txt lets GPTBot, ClaudeBot, PerplexityBot, etc. reach the site. Needs a GA4 property (for referral) and/or a site_url (for crawl coverage). Honest bound: AI referrals undercount badly (~70% land as Direct), and AI-Overview clicks count as Organic, not here.

Create or append many redirects at once via an account-level Bulk Redirect List (for site migrations). Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true AND a confirm value equal to list_name. Validates every item locally first and rejects the whole batch on any bad item (never half-applies); supports dry_run. Items are added asynchronously by Cloudflare.

Create one single (dynamic) redirect at the Cloudflare edge (e.g. a 301 for a renamed/migrated URL). Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true. Validates the target is reachable (no redirecting to a dead URL), refuses loops and duplicates, and supports dry_run to preview the rule first.

Delete one single (dynamic) redirect rule by its id (the rollback path for cf_create_redirect). Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true. Get the rule id from cf_list_redirects.

List DNS records for a zone (read-only). Useful for verifying the canonical host, CNAME flattening, and TXT verification records during SEO migrations.

List a zone's single (dynamic) redirect rules (source, target, status code, rule id) plus the account's Bulk Redirect lists (read-only). Call this before cf_create_redirect / cf_delete_redirect / cf_bulk_redirect_upsert so writes never clobber existing rules. Pairs with redirect_chain_audit and the migration_check prompt.

List the Cloudflare zones the API token can see, with status, plan, and id.

Get / configure / disable Cloudflare's managed robots.txt and Content-Signals policy on a zone (rides on the Bot Management config). action='get' reads the current state (read-only, un-gated). action='configure' enables/sets the managed robots.txt (managed_robots), the Content-Signals variant (cf_robots_variant: off / policy_only), and the AI-bot blocking levers (ai_bots_protection, content_bots_protection, crawler_protection). action='disable' turns managed robots.txt and the Content-Signals policy back off. NOTE: managed robots.txt and the Content-Signals policy are MUTUALLY EXCLUSIVE in Cloudflare - valid combinations are managed_robots=true + cf_robots_variant='off' (managed robots.txt) OR managed_robots=false + cf_robots_variant='policy_only' (the policy); the tool rejects the invalid combo locally. Writes are gated (SEO_MCP_ALLOW_DESTRUCTIVE) and need confirm=<zone>; supports dry_run. Always returns a caveat distinguishing the stated-preference signals from the levers that actually enforce at the edge.

Purge specific URLs from the Cloudflare cache so crawlers refetch updated content. Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true.

Purge the entire Cloudflare cache for a zone (affects all visitors). Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true AND a confirm value equal to the resolved zone hostname, so a full purge cannot fire ambiguously.

Audit a Cloudflare zone's settings for SEO and crawl/index safety (read-only): SSL/TLS mode, Always Use HTTPS, HSTS, Automatic HTTPS Rewrites, Brotli, and browser cache TTL. Findings are severity-graded with the reason and a benign exception each. Cloudflare cannot see the origin, so the SSL and Always-HTTPS checks are phrased as 'verify,' not 'fail,' and HSTS is never hard-failed (premature HSTS is dangerous and hard to undo). Answers 'are my edge settings silently sabotaging crawl/index?' It grades the edge layer only, not origin behavior.

Write the SEO/crawl/security Cloudflare zone settings that cf_settings_audit grades, to close the audit -> remediate loop. Gated: requires SEO_MCP_ALLOW_DESTRUCTIVE=true. Changing ssl_mode, or any HSTS change that raises protection, additionally requires confirm=<zone> (HSTS also requires acknowledge_hsts_risk=true). Validates locally first and supports dry_run (before/after preview). Needs the Zone Settings:Edit token scope (the audit only needs Read).

Read-only Cloudflare Web Analytics (edge RUM). With no argument, lists the account's Web Analytics sites; with host_or_tag, returns one site's detail. Edge-measured traffic complements GA4 (it sees visits GA4's JS tag may miss). Create/delete are intentionally not exposed.

Zone overview for a hostname: status, plan, paused, name servers, created/modified.

Fetch a URL and analyse its canonical link tag: report whether it is self-referential, cross-host, protocol-mismatched, or trailing-slash drift; flag missing canonical, and follow one canonical hop to confirm it itself returns 2xx.

Gather the hard evidence for a content brief: fetch the competitor pages and return their heading union, median word count (a FLOOR, not a target), schema types, and entity coverage, plus the 2026 GEO writing directives and validation rules. Pass competitor_urls (e.g. the top SERP results) for a true brief; with none it falls back to your own ranking pages via GSC (refresh-vs-new, NOT the full SERP). The host writes the brief prose from this; SEOMonster supplies rules + evidence. Read-only; no ranking guarantee. Backs the content_brief prompt.

Rank data-grounded content/blog topics from your own Search Console data. Fuses CTR-vs-expected gap, striking-distance position, demand volume, and momentum into a transparent opportunity score, flags cannibalization, and reports the click upside plus the score's components. Read-only. It prioritizes demand you already have; it does not do cold-start keyword research (that needs existing impressions) and does not write the content or guarantee a ranking.

Return the last 25 weekly collection periods of Core Web Vitals at p75 (LCP, INP, CLS, etc.) for a URL or origin via the Chrome UX Report History API. Complements psi_analyze with the trend axis.

Return the CURRENT (latest 28-day) p75 Core Web Vitals snapshot (LCP, INP, CLS, FCP, TTFB) for a URL or origin via the Chrome UX Report queryRecord API, with each metric's GOOD/NEEDS_IMPROVEMENT/POOR category and an overall verdict. Use crux_history for the 25-week trend. Small pages/origins legitimately have no field data; that returns no_data.

Landing pages with their conversions / key events over the last N days, broken down by channel group and sorted by conversions descending. Filtered to organic search by default (the SEO view) so you see which entry pages actually drive organic outcomes. Convenience wrapper over ga4_run_report.

Organic-search health over the last N days: window totals (sessions, engaged sessions, engagement rate, average session duration, conversions) plus a day-by-day trend of the same metrics. Two GA4 reports under the hood.

Run a GA4 report (Analytics Data API runReport) with arbitrary dimensions, metrics, a date range, an optional dimension filter, and ordering. The workhorse GA4 tool. Dates accept ISO (YYYY-MM-DD) or GA4 relatives like '28daysAgo' and 'today'.

Audit a GA4 property's configuration for SEO-measurement readiness (read-only): is a web data stream present, are key events / conversions defined, is data retention long enough for year-over-year analysis, are content-group custom dimensions set, and (via v1alpha) is enhanced measurement on and what the Google Signals state is. Findings are severity-graded with the reason and the benign exception for each. Answers 'can this property actually measure my organic outcomes?' It checks hygiene, not whether your events are the right business events.

Internal site-search terms over the last N days: what visitors typed into the on-site search box (GA4 'searchTerm'), by event count and sessions. A direct content-gap signal. If the property has no site-search data (no search box, or enhanced-measurement site search is off), the tool says so honestly rather than implying zero demand. Convenience wrapper over ga4_run_report.

Top landing pages by sessions over the last N days, with engagement rate and conversions. Filtered to organic search by default (the SEO view). Convenience wrapper over ga4_run_report.

Sessions, engaged sessions, and conversions broken down by default channel group over the last N days. Separates organic from paid / referral / direct at a glance. Convenience wrapper over ga4_run_report.

Inspect several URLs in one call (capped at 25). Returns a result per URL plus a list of per-URL failures. Rate-limit and transient errors are reported per URL without failing the whole batch.

Compare two equal-length time windows (current vs prior) and return per-key deltas in clicks, impressions, CTR, and position, plus keys present in only one window. v0.2.0: optional sort_by / sort_dir / min_delta_* filters and an anomalies_only z-score gate so the same tool covers 'biggest gainers', 'biggest losers', 'meaningful movers', and 'statistical outliers' without needing separate tools. When anomalies_only=true, the response's filters_applied.sigma_used field reports the population-stdev (sigma) actually computed from the matched rows' sort_by metric distribution; the effective z-score cutoff applied is `sigma_threshold * sigma_used`.

Heuristic coverage audit. The GSC Index Coverage report is NOT exposed in the public API, so this tool takes a user-supplied URL list (typically pulled from a sitemap) and bulk-inspects each, then rolls up the verdicts (PASS / PARTIAL / FAIL plus coverage_state frequencies). Honest substitute for the Coverage UI; the AI host can use the rollup to decide which URLs to deep-inspect.

Pages whose impressions fell most over the last N days vs the prior N days. Equivalent to `gsc_compare_periods` with `dimensions=["page"]`, `sort_by="delta_impressions"`, `sort_dir="asc"`. Use this for lifecycle triage (find pages to invest in or rescue).

Inspect a single URL via the URL Inspection API: index verdict, coverage state, crawl info, canonicals, and mobile/rich-results summaries when present.

Ground LLM-brainstormed keyword candidates against your own Search Console data to find net-new terms you have no current footprint on. YOU (the host) supply `candidates` -- brainstorm them from the site's winning queries first (call gsc_top_queries). For each candidate this returns a footprint verdict (covered / thin / none) and a confidence band from sibling-strength (does it share tokens with queries that already win impressions). 'none' means no VISIBLE footprint, not no demand (GSC hides ~75% of impressions); net-new terms are scored hypotheses, not facts. Read-only.

List every Search Console property the configured credentials can see, with each property's permission level. Use this to discover valid site_url values.

List the sitemaps Google knows about for a property, with submission and indexing status.

Queries that have impressions in the current window but had no impressions in the prior comparison window. Useful for spotting emerging topics or pages that just started ranking.

Multi-property fleet view. Lists every property the credentials can see and returns a one-row summary per property (total clicks, impressions, average CTR, average position) for the last N days. The single fastest answer to 'how is the portfolio doing?' across agency or multi-brand setups. Honors the v0.2.0 alias convention (`days`) and applies the configured data_state.

Queries that drive impressions but very few clicks. These are content opportunity signals: a page is being shown for the query, the click experience is not landing. Returns rows sorted by impressions desc.

Queries already ranking in the top N positions but with below-target CTR. These are title and meta-description optimization candidates. Returns rows sorted by impressions desc (the bigger the impression volume, the more clicks a CTR lift unlocks).

Ask Google to (re)crawl one or more URLs via the Indexing API (URL_UPDATED). Requires the indexing scope. Available by default. Accepts a single `url` (string) or `urls` (array up to 100). A scope or disabled-API error stops the batch and is returned with remediation; per-URL errors are collected. `notify_time: null` on success is normal upstream behavior, not a failure.

Query Search Console search analytics (clicks, impressions, CTR, position) for a property, with dimensions, a date range, optional dimension filters, and data_state. The workhorse GSC tool.

Submit a sitemap to Search Console. Requires the writable webmasters scope. Available by default (a routine SEO task; not gated). Accepts either `sitemap_url` (preferred, friendly name) or `feedpath` (the raw API field name; kept for back-compat).

Convenience wrapper: top pages for a property over the last N days by clicks. Equivalent to gsc_search_analytics with dimensions=["page"]. Uses the configured data_state (default 'all'; 'final' lags 2-3 days behind the dashboard).

Which pages rank for a specific query. The classic cannibalization audit input: if multiple pages rank for the same query, you typically want to consolidate or differentiate them. Filters search_analytics by an exact query match and returns rows by page dimension.

Convenience wrapper: top queries for a property over the last N days by clicks. Equivalent to gsc_search_analytics with dimensions=["query"]. Uses the configured data_state (default 'all'; 'final' lags 2-3 days behind the dashboard).

Pages whose impressions grew most over the last N days vs the prior N days. Equivalent to `gsc_compare_periods` with `dimensions=["page"]`, `sort_by="delta_impressions"`, `sort_dir="desc"`. Use this for lifecycle triage (find pages to invest in or rescue).

Validate hreflang link tags across a set of URLs. Checks: missing reciprocity, broken target URLs, duplicate hreflang on one page, missing self-link, missing x-default when there are 3+ language variants. Returns per-URL findings and a global findings list.

Submit multiple URLs to IndexNow in one batched request. All URLs must share the same host (IndexNow rejects mixed-host batches with HTTP 422). Cap: 10000 URLs per call (IndexNow's documented limit).

Submit a single URL to IndexNow (Bing + Yandex + Naver + Seznam + Yep). Complements gsc_request_indexing for the non-Google engines. Requires SEO_MCP_INDEXNOW_KEY plus a key-verification file at https://<host>/<key>.txt.