SEO Audit Console

Full dashboard dataset for the UI widget. App-only — not for direct use.

Update just the GSC URL Inspection data (coverage, indexing, Google-vs-declared canonical, last crawl) for top URLs into url_inspection. Quota-limited — samples top pages by clicks. Async job. For a full refresh use refresh_property. Grain: one row per inspected URL. Joins: url_key → pages/GSC.

Demand-first keyword clustering from stored data - NO paid calls. Give it a keyword list (or omit keywords to use your top 500 GSC queries) and it clusters them by the page Google ALREADY answers them with (two keywords that rank via the same URL belong together - the strongest clustering signal, free from your own GSC data), then groups non-ranking keywords lexically. Each cluster gets a deterministic verdict: OWN (best position <=3), WEAK (4-20), ABSENT (no ranking page), plus the ranking URL, summed 90-day impressions and clicks. The keyword-research workhorse: paste a client keyword list, …

[Paid: Keywords API, cheap, cached 20d | Use for: demand sizing] True monthly search volume + CPC + competition for keywords (DataForSEO KEYWORDS_DATA). Served from a 20-day cache; live calls are serialised. Default location: United States (2840).

[Paid: Backlinks subscription (separate - 40204 = not activated), one cached call ~$0.024 | Use for: prospect list for link outreach] Answers "what links do our competitors have that we don’t?" - and for a SINGLE company, "what links does company X have that we don’t?" (pass competitors:["companyx.com"]). One DataForSEO domain_intersection call over the target set (excluding your domain), aggregated per prospect domain: how many of the targets it links to, its DataForSEO domain trust (rank 0-1000), worst spam score, whether the link is followed, the anchor/link-type mix. Default sort is the li…

List the technical-SEO check categories this server can evaluate.

List Google Search Console properties accessible to the service account.

Cluster the crawled pages into templates (by URL morphology + JSON-LD @type) and return each template with its page count and a representative exemplar URL. A single template-level fix corrects every page in the cluster — this is the map for per-template analysis. Needs a crawl (run refresh_property / start_crawl first).

[Paid: Labs, one cached call per domain (<=5) | Use for: sizing the organic market and who owns it] Build the organic market map: your domain plus up to 4 named competitors, ONE cached Labs ranked_keywords pull each, unioned into a keyword universe. Returns total monthly demand (deduplicated search volume), each domain's share of voice (ETV share) overall and per topic cluster, and the leader per cluster - the "here is the market, here is who owns it, here is where to attack" table that opens an engagement. Deterministic: the competitor set and ranked keywords (Labs index). Judgement: ETV is D…

[Google News is FREE (no key); DataForSEO adds a paid, richer Google News SERP | Use for: what has been PUBLISHED on a topic — freshness, "what changed since {date}", competitor coverage] Recent news for a keyword from Google News (the free RSS search feed — Google's old News API is retired) and/or the DataForSEO Google News SERP. Returns title, source, publish time, URL. Default source "both" merges + dedupes; pass source:"google" for a free keyless lookup, "dataforseo" for the paid SERP only. Default location: United States.

Return the canonical url_key used to join GSC data to crawl data.

Return a link that opens the property's interactive dashboard in the browser, starting the localhost dashboard server if it isn't running. Offer this after a sync, crawl, refresh or audit finishes. Where no local port can be served (e.g. the Docker MCP gateway), it says to use get_dashboard, which renders the dashboard in chat.

[Paid: Labs, cheap, cached 20d | Use for: page-level content gap vs named rival URLs] Find keywords that competitor pages rank for but your page does NOT (the content gap). Pass competitor URLs as `competitorUrls` (max 20; wildcards like https://site.com/blog/* allowed) and your own URL(s) in `excludePages` (max 10) to subtract. Returns gap keywords sorted by search volume, with each competitor’s rank. Labs call, cached 20 days. Pass location + language.

[Paid: On-Page, ONE URL per call | Use for: CWV evidence on a page that earns clicks] Run a live Lighthouse audit for ONE url: lab Core Web Vitals (LCP, CLS, TBT, FCP, Speed Index), category scores (performance/SEO/best-practices/accessibility), and the top time-saving opportunities. Complements GSC/CrUX field data (aggregate + delayed) with on-demand lab data. Pass siteUrl to persist the CWV so run_audit can surface high-yield-cwv-fail. Paid On-Page call (~2000 credits), slow (~20–120s), cached 20 days.

Fetch arbitrary URLs on the property's own host (apex or www) — including ones you generate to test a hypothesis — and report per URL: status, the full redirect chain and final URL, canonical, meta robots and X-Robots-Tag, title, a content fingerprint (equal contentHash = same page), and cache headers (cf-cache-status, x-cache, age, cache-control). The crawler only fetches what it discovers; use this to test case, trailing-slash, parameter, scheme, host or typo variants, redirect rules, or whether a 404 in the data still 404s. `variantsOf` (up to 6 URLs) is a preset: it generates uppercase, ti…

[Paid: Backlinks subscription (separate - 40204 = not activated), cached 20d | Use for: authority + dead-backlink recovery] Fetch the property’s backlink profile (overall summary — total backlinks, referring domains, Domain Rank, broken backlinks/pages, nofollow share — plus per-page backlink/referring-domain counts) into page_backlinks, and resolve each backlinked page’s live HTTP status so run_audit can flag external backlinks pointing to dead (4xx/5xx) pages. Also collects the profile behind the dashboard's Links tab (#47): referring domains (live and lost, with rank, spam score, first seen…

Run a single named check (see list_checks) and return its affected URLs + evidence. Token discipline: `columns` narrows evidence to the named keys, string evidence values are truncated at 120 chars (trailing …), `offset` + `limit` page through large result sets (the footer states showing X–Y of TOTAL). Grain: finding per URL with evidence JSON. Joins: url_key across all sources.

Direct read-only access to the property database — the composition layer's escape hatch. AGGREGATE IN THE DATABASE and return answers, not rows: default mode groups by the columns you name and returns counts, percentages and sum/avg/min/max metrics (classic crawler-style distributions in one call — "status codes by folder", "clicks by page_key"). mode:rows returns raw rows with strict token discipline (curated default columns, 120-char cells, loud paging footer). Filters are exact/range/like/in/match and always parameterised; op:"match" is full-text search via FTS5 (words, "phrases", prefix*, …

[Paid: Labs, cheap, cached 20d - ONE call replaces any keyword loop | Use for: full keyword footprint of a domain/page/folder] Every keyword a target ranks for in Google organic — scope it to a whole domain, a subdomain, ONE page (scope:url with the full URL), or a subfolder (scope:folder + folder:"/blog/"). Returns keyword, position, search volume, ETV, the ranking URL, plus keyword difficulty, search intent and SERP features where the response carries them. Set aioOnly:true to list only keywords whose SERP shows a Google AI Overview (your AIO exposure list; per-keyword CITATION checking uses…

[Paid: DataForSEO SERP per page (~$0.004 each, plus a small refundable surcharge for loading async AI Overviews), bounded to the batch | Use for: the deep "why are we behind and what to add" recon] For each of your striking-distance pages (auto-selected: the page's top query at positions 3-15, highest impressions first; or pass explicit urls), this fetches OUR live page with the crawler, pulls the live Google SERP (DataForSEO SERP-advanced, depth 20 organic), and classifies WHY we are behind using the organic-rank x AI-Overview-citation matrix: defend-and-deepen (cited + strong), accuracy-or-f…

The content-recon to-do board. With no id: list to-dos (optionally filter by page or status), grouped by page with each page's verdict — the pick-a-page-to-work-on surface, and the hand-off to content-machine. With id: update one to-do — set status (open → researching → drafted → shipped → dismissed) and/or append a dated annotation note (your own observations, the history). On status:shipped with remeasure:true it re-fetches the SERP and records the outcome, so you can see whether the fix moved you from AIO-uncited to cited, or up the organic ranks.

Full refresh for a property in one async job: GSC sync → site crawl → URL inspection → DataForSEO rank history → CrUX field data (when CRUX_API_KEY is set; each target at most once per 7 days). Opens a live progress widget (phases + counts). Set gsc/crawl/inspect/ranks/crux=false to run just part. GSC sync is "lite" (date×query×page) by default — set segments=true to also pull device/country breakdowns (much heavier on large sites). GSC sync is INCREMENTAL: the first sync pulls the window (default last 90 days; pass startDate to go deeper), later syncs only fetch new days — set full=true to fo…

[Paid: SERP call PER KEYWORD - only for clicked/explicit keywords, never lists | Use for: the SERP context of a single keyword] People Also Ask questions and related searches for a keyword (DataForSEO SERP). Powers click-through "related terms" on the keyword charts. SERP call — cached 20 days.

[Opt-in: needs the optional `playwright` package + `npx playwright install chromium`] Render a BOUNDED, sampled set of indexable pages in headless Chromium and diff the rendered DOM against the raw HTML the crawler fetched — the "Response vs Render" check the classic desktop crawlers are known for. Surfaces content and internal links that exist ONLY after JavaScript runs (what a first-pass crawl, and Google before it renders, can miss). Persists to render_audit (raw vs rendered canonical, title, robots meta, X-Robots-Tag, noindex and main-content words); run_audit then reports `content-only-in…

Heuristically resolve each top page’s primary entity (by H1, fallback title) to a Wikidata QID, and store subclass-of / part-of relationships between them. Unlocks the entity-internal-link-gap finding (suggests internal links a topical mesh implies). Heuristic — those findings are judgement (N), shown only with includeJudgement. Free (public Wikidata API), cached ~45 days, on-demand. Async job — poll check_sync_status.

Run the technical-SEO checks against synced data and return scored findings, ranked by expected clicks per dev-hour — Priority = (T × Y × C) / E (T = clicks at stake from real GSC data, Y = expected yield, C = certainty, E = effort hours). Crawl + GSC + URL-inspection checks. Set includeJudgement=true to include heuristic (N) checks. Returns the counts (byCheck, bySeverity, byCategory, notCollected) and the top findings (topLimit, default 10, max 50) with trimmed evidence; each check's title and fix appear once in `checks`. query_audit is the drill-down: every finding of one check, full eviden…

Write content-recon findings back into the trackable ledger for a page — the gaps and originality the research session found by diffing the competitor pages and ranking videos against our content. Each to-do is an action with a type (content-gap / originality / schema / freshness / format), rationale and evidence. Snapshots the page baseline so the fix's effect on rank/AIO-citation is measurable later. Run recon_targets first (it classifies the page and seeds the deterministic to-dos); this adds the judgement ones. Track everything with recon_todos.

Run a local cross-encoder reranker (ms-marco-MiniLM-L-6-v2, downloaded once to a cache, no Python) over each ranking page's heading chunks against its top Search Console query, and persist the single best-passage relevance score. It SCORES relevance (a classifier — it cannot hallucinate). Powers the `weak-passage-answer` check — the "RAG snippetability" test: does any passage confidently answer the query, the way AI/passage search re-ranks? Returns the score for EVERY page it scored (not just the failures) with the flag threshold, the band convention and the run's own distribution, so a passin…

[Paid: Labs, cheap, cached 20d, up to 1000 kw/call | Use for: intent mapping before content planning] Classify the search intent (informational / navigational / commercial / transactional) of keywords — primary label + probability + secondary intents. Use it to spot intent mismatch: e.g. a transactional product page ranking for an informational query (a common cause of high impressions / low CTR). Pass siteUrl to persist the intents so run_audit can surface intent-vs-pagetype-mismatch. Labs call, cached 20 days, up to 1000 keywords. Language-only (no location).

Overview of every tool/feature with an example prompt for each. Start here.

[Paid: Labs, ONE cached call | Use for: AI-Overview / zero-click exposure with real numbers] How much of your keyword universe carries each SERP feature - AI Overviews, featured snippets, People Also Ask, shopping, video - weighted by search volume, and how much of that volume you already rank page 1 for. ONE DataForSEO Labs ranked_keywords pull (top-volume sample, cached 20 days, never per-keyword SERP loops). Answers "how exposed are we to AI Overviews / zero-click?" with real numbers. Deterministic: feature PRESENCE (from the Labs index). Judgement proxy: page-1 rank stands in for feature o…

Start a localhost-only webserver and return a URL that opens the full interactive dashboard in your browser — live data straight from the local database (always current, unlike export_report snapshots), a property switcher, working CSV downloads, and no host widget limits. The server stays up while the MCP server runs; call again with stop=true to shut it down. Localhost only — nothing is exposed to the network.

Update just the crawl: fetch the site into the local database (async job — poll with check_crawl_status every 60-120 s; it reports the stage, pages crawled vs maxPages, pages/sec and an ETA). Large sites: maxPages up to 50000 (default 4000); a 10k-page crawl is typically 20-40 minutes and streams pages/links to SQLite as it goes, then runs the post-crawl stages (link graph, snapshot, outbound-link check). Rate limits are adapted to (#54): a 429 (or 503 with Retry-After) halves the effective concurrency and honours Retry-After, 20 clean responses add one back up to maxConcurrency, and rate-limi…

Propose NEW pages grounded in real Search Console demand: queries you already get impressions for but rank 11+ (no winning page), after subtracting demand you already satisfy (you rank ≤10, or a page already covers the query, or one URL owns it). Survivors are clustered into one proposed page per intent and scored by impressions × intent. Returns evidence + the nearest existing page to link the new page from. Computable from synced GSC + crawl — no paid calls (run search_intent siteUrl:<property> first to weight by intent).

Description is built at runtime.

Update just the GSC search-analytics history (async job — poll with check_sync_status). Incremental: the first sync pulls the range (default last 90 days; pass startDate to go deeper), later syncs only fetch new days — set full=true to force a full re-pull. Lite by default (date×query×page); set segments=true (or pass dimensions) to also pull device/country. For a full refresh use refresh_property. Grain: date × query × page. Joins: page_key (url_key) → crawl/inspection/backlinks; query → DataForSEO keyword tools.

[Paid: Labs, cheap, cached 20d | Use for: where rival traffic actually lands] The top organic pages of ANY domain or subdomain, ranked by estimated traffic value (ETV): page, ranking-keyword count, ETV and top-3 / top-10 keyword counts. The Semrush-style "top pages" view — works on competitors, no crawl or GSC needed. Labs call, cached 20 days. Pass location for the right market. Grain: page × domain. Joins: domain → Labs tools; URL → pages.url_key on your own property.

[Paid: Labs, <=4 cached calls | Use for: topic-level gap plan vs competitors] What related topics should this site cover to be seen as expert in its space? Pulls competitor keyword footprints (DataForSEO Labs ranked_keywords, ≤4 bounded + 20-day-cached calls), subtracts everything YOU already surface for (every GSC query with impressions) or already have a page about (title/H1/slug near-match), clusters the surviving gap keywords lexically, and scores each topic by summed search volume × how many competitors rank there. Each topic names the owning competitor with an example URL, your nearest e…

[Paid: Keywords API, cheap, cached 20d | Use for: seasonality + "is this rising or fading" — should we update now, when to publish; category monitoring — is a whole market growing] Google Trends relative interest (0–100) over time (DataForSEO KEYWORDS_DATA / google_trends). Give up to 5 keywords, a categoryCode (find it with trend_categories), or both: keywords alone search all categories; keywords + categoryCode narrow a term to one industry (e.g. "jaguar" in Autos); categoryCode alone returns interest in the WHOLE category, no keyword needed. related:true adds the top + rising related topics…

Ingest the DataForSEO over-time sequence (monthly rank distribution + ETV) into rank_history, so rank charts have a real time axis reconciled with GSC. DataForSEO Labs call — cached 20 days. Pass location once (e.g. "Australia", "United Kingdom") — it is saved per property. Async job.

[Free endpoint; needs DataForSEO credentials | Use before topic_trend with categoryCode] Search the Google Trends category tree (~1,400 categories) by name and get the codes topic_trend takes, with each match's parent so you can pick the right level. Omit query to list the top-level categories.

Turn the crawl's 404/410 list into a short redirect map. A GSC-seeded crawl reports every URL Google has seen, so real lost pages (traffic, backlinks) sit among many mangled variants of real slugs from scraped or AI-rewritten links. Each 404 gets its discovery sources (internal link / sitemap / GSC / backlink), clicks, impressions and last-seen date from search_analytics, backlinks and referring domains, and the closest live indexable URL by slug similarity (token overlap or edit distance, 0-1). Buckets: redirect (has clicks, ≥minImpressions impressions or referring domains, AND a match ≥minSc…

[Paid: SERP call PER KEYWORD, cached 20d | Use for: finding the videos that rank for a topic — content research, competitor video scan, freshness] The ranking YouTube videos for a keyword (DataForSEO YouTube organic SERP). Returns title, URL, channel, view count, publish date, duration and shorts/live flags, in rank order. Watch the top few to see what the winning videos actually cover. SERP scope, 20-day cache. Default location: United States (2840).