SEO tools: the paid lists, Reserve and vendor notes (prices checked 2026-10-24)

Read this only when a paid budget or credential is set, the user asks for a paid tool, or a free row’s Reserve rule may apply. The default free stack, the labels and the rules behind every list are in tools.md. Facts new to these lists have their sources in tools-sources.md.

Paid lists by category

Each category has a small and a big paid list and a Reserve table. They take over from the default free stack per category as tools.md describes.

Keywords and competitors

For each market-language pair: what people search for, how often (volume and seasonality), how hard it is to rank, the intent and questions behind it, and which competitor domains win those searches with which keywords. The run clusters and prioritises the keywords each page should own and writes them to the research store (seo-research/).

Georgia is a special case. Google Ads has no Georgian language criterion, so GE-ka and GE-ru keywords go into the same location-only tasks. DataForSEO Labs has no Georgia location, so Georgian competitor databases come from SE Ranking.

Small sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Research store (research_store.py)	Main	The reuse layer. status lists stale or missing datasets per locale, so paid calls cover only the gaps. Each value keeps its source and date, so values from two sources sit side by side for the cross-checks	Local script, no auth, runs in every scheduler. Skills write only through ingest, set-status and glossary add. Semrush values stay within 30 days and out of commits	$0
2	Search Console API, Search Analytics (through claude-seo)	Main	The only measured demand: impressions, clicks and position per query, page and country, in any language. The free check on every vendor volume. It has no language dimension, so ka, ru and en come from the URL path or the query’s script	claude_seo.py gsc-export with the property named, on the user’s one-time OAuth sign-in (webmasters.readonly) or a service account an owner adds as a property user. 25,000 rows a request, 50,000 rows a day per search type, 2–3 days behind. Without either: exports, or Search Console in the browser while the user is signed in	$0
3	DataForSEO Keywords Data: Google Ads (search_volume, keywords_for_site)	Main	Google’s own Ads volumes, with 12 monthly values for seasonality. Exact for Georgia (2268) and its cities. keywords_for_site gives Google’s ideas for a competitor’s domain, the cheapest competitor-topic source in Georgia	Connector, Standard queue (task_post, free tasks_ready, task_get), priced by the guard. Keywords up to 80 characters and 10 words	$0.06 a task of up to 1,000 keywords. A quarterly shortlist plus 3 competitors ≈ $0.10
4	DataForSEO SERP API (Google organic, People Also Ask)	Main	The observed top 10 per keyword and locale: competitor domains and their share of top-10 slots, SERP features and PAA questions. serp_cluster.py groups keywords by overlap. Where vendor databases are thin, as in Georgian, this is the most trustworthy competitor evidence	Same connector, Standard queue. ingest serps writes competitors and questions to the store	$0.0006 a SERP. 300 keywords × 2 pairs ≈ $0.36 a snapshot, $0 for keywords rank tracking already holds. ≈ $0–0.80
5	DataForSEO Google Autocomplete API	Main	Idea expansion through an official API with exact location. It replaces the unofficial autocomplete endpoints as the main idea source	Same connector, Standard queue, cluster-head seeds only	$0.0006 a task. 30 seeds × ~40 expansions × 2 pairs ≈ $1.44 a quarter, ≈ $0.50
6	SE Ranking Data API (to build: guard prices)	Main	The cheapest competitor keyword lists and gaps, with a Georgia database (ge). domain/keywords, domain/keywords/comparison and domain/competitors cost 100 credits a request for up to 1,000 rows. Records carry volume, difficulty, intent and a 12-month trend	Data API key or MCP (https://api.seranking.com/mcp), reads only, server named seranking. Until the guard has its prices it allows 5 calls a run, so competitors come from rank 4 meanwhile	≈ $0.20–0.60 (1,000–3,000 credits). A $50 wallet up front, which never expires
7	compare_pages.py	Main	Compares our page with the pages that win its keyword: size, structure, schema, trust signals, and the phrases competitors share that our page never mentions	Plugin script, no auth. Competitor URLs come from the stored SERPs	$0
8	DataForSEO Labs (ranked_keywords, domain_intersection, bulk_keyword_difficulty) (guard: rows)	Cross-check	Compares a competitor’s keywords and positions with SE Ranking for the same domain and market, and Labs difficulty with SE Ranking’s and our SERP-based winnability. Worth it quarterly in markets Labs covers, when the gap list will create pages. No Georgia	Same connector, Live only, always with limit. As shipped the guard clamps limit to max_rows (50)	≈ $0.13 per covered pair. $0 for a GE-only site
9	keyword_ideas.py, Bing suggestions	Cross-check	Compares Bing suggestions with DataForSEO’s Google suggestions for the same seeds. Only Bing is a different source: the script’s Google and YouTube modes call the same Google service. A phrase both engines suggest ranks higher. Free, so run it on cluster heads each quarter	No auth. Unofficial, undocumented endpoints, a deliberate exception: --delay 0.25 s or more, modest volumes	$0
10	Competitor sitemap inventory (to build: polite mode)	Cross-check	Compares what competitors publish (URL templates, new sections by first-seen date) with the keyword lists from SE Ranking, Labs or keywords_for_site. A section with no vendor keywords points to a gap in the vendor’s database, common in ka and ru. Confirm it with a SERP check. Free, so run it on every monthly gap pass	Until a polite mode exists (robots.txt, one request at a time per host, stop on 403, 429 or a challenge, snapshot diffs): sitemap_check.py --sample 0 --urls-out on competitors, diffed outside the script. Store URLs only, never page text	$0
11	WebSearch SERP samples (seo-serp-researcher agent)	Cross-check	Compares the top-10 domains and the main page type for 20–50 head terms with DataForSEO’s SERP. If both agree, act. If not, re-check with DataForSEO	Claude’s own WebSearch and WebFetch, allowed in --allowedTools. Tag [measured: WebSearch sample]	$0 (session tokens)
12	Wayback Machine CDX API	Cross-check	Compares first-seen dates from our sitemap snapshots with the archive’s first capture of the same competitor URLs, and backfills history before our first snapshot. Run it when a competitor section appears or a competitor gains	Free public API with no key (web.archive.org/cdx/search/cdx?url=…&output=json), at a polite rate	$0

Big sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Research store	Main	As Small. Its TTLs keep paid spend proportional to change, not to site size: volumes and SERPs refresh after 30 days, competitors after 90	As Small	$0
2	Search Console API (through claude-seo)	Main	As Small, for the freshest days	As Small	$0
3	Search Console bulk export to BigQuery	Main	No row cap, and anonymised-query totals: the long tail and per-language splits that the API truncates. It doesn’t backfill, so turn it on at onboarding	An owner turns it on once (needs Cloud billing). Runs read it through Google’s official BigQuery MCP (https://bigquery.googleapis.com/mcp) with only execute_sql_readonly exposed. At most 3,000 rows and 3 minutes a call, so aggregate in SQL	$0 within BigQuery’s free tier (10 GiB stored, 1 TiB of queries), then $6.25 a TiB. Usually $0–10
4	DataForSEO Google Ads	Main	As Small, in bulk	As Small, up to 100 tasks per POST	20,000 keywords a pair monthly is 120 tasks ≈ $7.20, plus keywords_for_site for 10 competitors in 4 locations ≈ $2.40. ≈ $10
5	DataForSEO SERP API	Main	As Small, for thousands of keywords per locale. It drives share of voice and SERP-overlap clustering	As Small	5,000 keywords × 6 pairs ≈ $18, less the keywords rank tracking already holds
6	SE Ranking Data API (to build: guard prices)	Main	The competitor-keyword database for the GE pairs and any market Labs lacks. Cheap per row: 1,000 rows for 100 credits	As Small	10 competitors × ge × 10 pages of 1,000 rows ≈ 10,000 credits ($2) a quarterly pull. ≈ $1
7	DataForSEO Labs (ranked_keywords, domain_intersection, competitors_domain, bulk difficulty and traffic) (guard: rows)	Main	The cheapest bulk competitor-keyword and traffic-estimate database where it has data: 94 countries, here GB, IL and DE. No Georgia	Same connector, Live only, explicit limit. The free dataforseo_labs/status endpoint shows database freshness	10 competitors × 3 covered pairs × 5,000 keywords ≈ $19.80 a quarterly pull, plus bulk difficulty ≈ $2.40 a refresh. ≈ $7.40. Under the current 50-row clamp a pull costs ≈ $54
8	DataForSEO Google Autocomplete API	Main	As Small, for thousands of seeds	As Small	200 seeds × ~48 expansions × 6 pairs ≈ $35 a full expansion, run quarterly. ≈ $12
9	compare_pages.py	Main	As Small	As Small	$0
10	SE Ranking Data API in Labs markets (to build: guard prices)	Cross-check	Compares a competitor’s keyword set and positions with Labs ranked_keywords for the same domain and market. A keyword both report enters the gap list. A keyword in only one gets a SERP check first. Worth it on every quarterly pull, because it costs almost nothing	As Small	≈ $6 a pull for 3 databases. ≈ $2
11	Ahrefs API v3 (to build: REST client)	Cross-check	A fourth database, with clickstream-based volume, its own difficulty and top pages by traffic. Compares head-term volume and difficulty with DataForSEO’s Google Ads volume and SE Ranking’s difficulty, and a competitor’s top pages with Labs. A gap of more than about 2× on a head term [heuristic] is settled by GSC impressions or a SERP check. Georgia only as ka-GE. Worth it for decisions about new sections or markets	REST v3 with an API key, never the MCP (its terms forbid scripts). The hook doesn’t see REST calls, so the client checks the free subscription-info/limits-and-usage and the guard’s ledger before each run. 10-unit fields only on shortlists	$0 extra on the Ahrefs Standard plan priced under Backlinks. Dropped if that plan is dropped
12	DataForSEO Clickstream (bulk_search_volume) (guard: price)	Cross-check	Clickstream volume is measured differently from Google Ads, which merges close variants and rounds. Compares head-term volumes. A large gap flags merged variants or a seasonal spike. Only when a covered market is a target: Georgia coverage is [UNVERIFIED]	Same connector, Live only. The guard has no price for it, so it counts each call at 5% of the DataForSEO budget	2,000 head terms × 3 covered pairs ≈ $0.75
13	CrUX popularity rank (BigQuery)	Cross-check	Google’s own popularity band per origin, from Chrome navigations. Compares each competitor’s band with vendor traffic estimates. A vendor saying “larger than us” against a lower band makes the estimate suspect. Free, so run it each quarterly competitor review	BigQuery with the service account used for the Search Console export. Bands by country come from rank in the materialized country_summary table	$0 within the free tier
14	keyword_ideas.py, Bing suggestions	Cross-check	As Small	As Small	$0
15	Competitor sitemap inventory (to build: polite mode)	Cross-check	As Small, and the publishing rate per section. Confirm templates with no ranked keywords by a SERP check before sending them to seo-programmatic	As Small	$0
16	WebSearch SERP samples (seo-serp-researcher agent)	Cross-check	As Small, on the head terms of each pair	As Small	$0 (session tokens)
17	Wayback Machine CDX API	Cross-check	As Small	As Small	$0

Reserve

Tool	Use when …
Semrush (official MCP; Starter $199 or Pro+ $299, 50,000 units)	The owner already pays Starter or Pro+, and a scheduled claude -p run has kept a working sign-in for two weeks, or Starter is shown to issue an MCP API key (both unverified). Then read up to 40% of its units, with fixed row limits, as a third Georgian database beside SE Ranking. For a small site, also: SE Ranking returns fewer than 100 keywords for a competitor that holds 5 or more of our tracked top-10 slots. Not Main: a fixed $199 for data that costs a few dollars pay-per-use, and its unattended sign-in is unproven. The $139 SEO plan lists no MCP access
Ahrefs API v3 (small sites)	Ahrefs Lite or higher is already paid for links and the REST client exists. Then spend a few units on difficulty and a competitor’s top pages. Its MCP is for sessions someone is watching
Google Keyword Planner (UI export, or the Ads API)	The client runs Google Ads with sustained spend and wants first-party numbers without a reseller, or DataForSEO is unavailable. Not a cross-check: DataForSEO resells the same Google Ads data
Google Trends (API alpha)	Alpha access is granted, for consistently scaled 5-year interest by country and sub-region. Never scrape trends.google.com. DataForSEO’s 12 monthly values already cover seasonality
Keywords Everywhere (API and MCP)	The markets are among its public checker’s countries (US, UK, CA, AU, IN, NZ, ZA) and the team already uses its extension, for spot checks in watched sessions. Its volumes come from Keyword Planner, so it is not an independent cross-check. No Georgia. Credits expire after a year
AlsoAsked (Pro: API and MCP)	A question hub of 100 or more pages is planned, and one test query shows it returns Georgian or Russian questions for Georgia (repeats with fresh=false are free). Take Pro ($59) for that month. It re-measures Google PAA, which DataForSEO already collects. The guard counts it in the shared “other” bucket
AnswerThePublic	A person brainstorms question phrases in a watched session and saves them as seeds. No API or MCP
SpyFu (API on Pro + AI or Team)	The work needs a competitor’s historical Google Ads keywords and ad copy in one of SpyFu’s 16 countries. No Georgia or Israel
Moz API	A stakeholder reports Moz keyword metrics or Domain Authority. Free tier 50 rows a month; Starter Medium $20 for 3,000 rows with no overage. Its official MCP is in beta on paid plans. It adds no locale the Main tools lack
Ahrefs Free (UI)	A person wants Ahrefs’ view of the verified site’s top 1,000 keywords. UI only
Semrush Traffic & Market (.Trends API)	An enterprise client needs Semrush’s panel traffic and audience data instead of Similarweb: $289 a month per user (annual) plus a separate quarterly Trends API subscription
Similarweb (MCP; API-only, Business or Enterprise plan)	A board-level market-sizing decision is due in a market Similarweb covers and an API contract is budgeted (quote-only; third parties report $15,000–40,000 a year [UNVERIFIED]). Then compare its visits with Labs and SE Ranking estimates as an order of magnitude only. The guard counts it in the shared “other” bucket
Bing Webmaster Tools keyword API (GetKeywordStats, GetRelatedKeywords) (paid lists)	Bing has 10% or more of the site’s search clicks (Bing Webmaster clicks against Bing plus Search Console clicks), for a free second engine’s demand numbers. No Bing client is built yet, and Georgian support is unverified. In the default free stack it is a Cross-check
Google Ads MCP (official, read-only)	The site runs Google Ads: its search terms report gives real queries with impressions. It has no Keyword Planner
Search Console bulk export (small sites)	Search Console API pulls hit 50,000 rows a day per search type. In the default free stack it is Main
DataForSEO Clickstream (small sites)	A covered market is a target and one volume decides whether a page gets built
YouTube Studio, vidIQ, TubeBuddy	A YouTube channel is in scope. A person reads them; no run connects
keyword_ideas.py, Google and YouTube suggestions (paid lists)	Google: DataForSEO Autocomplete is unavailable for a run (its budget is used up, or an outage). In the default free stack the Google mode is a Cross-check against Search Console queries. YouTube, in every list: a YouTube channel is in scope and video-search phrases are wanted, which no other tool here gives. Both modes call the same unofficial Google endpoint, so keep --delay at 0.25 s or more
CrUX popularity rank (small sites)	A Cloud project with billing already exists, for example for the Search Console export, and a competitor’s vendor traffic estimate decides its priority. Then compare its band with SE Ranking’s estimate, as in the big list. In the default free stack it is Main

Backlinks and authority

Who links to us and to our competitors, which links were gained or lost this week, how strong and clean the linking sites are, and which domains link to competitors but not to us. seo-links acts on it: reclamation redirects through the autonomy levels, a prospect list, and the disavow decision, which stays gated at every level.

Small sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	DataForSEO Backlinks API	Main	The only backlink source already wired end to end: connector, guard prices and the weekly B5 delta pipeline. Own and competitor summaries (including broken backlinks and broken pages), referring domains, anchors, and new and lost links. domain_intersection takes 20 targets and 10 excluded, so “links to competitors but not to us” is one call. Bulk rank and spam score take 1,000 targets a request. Its depth in .ge and Russian-language sites is unmeasured [UNVERIFIED]	Connector with the login in the keychain. Live only, 30 simultaneous requests. A dashboard daily spend limit makes calls fail (40203) rather than overspend. As shipped the guard clamps limit to max_rows (50)	Own site + 3 competitors: weekly deltas ≤ $1.39, a quarterly gap ≈ $0.03, a 1,000-target prospect score $0.06. ≈ $1.50
2	Bing Webmaster Tools API (GetLinkCounts, GetUrlLinks) (to build: Bing client)	Main	Free first-party links from a search engine’s own crawl, independent of every vendor, with the source URL and anchor for each linked page. The free floor when paid budgets are 0. A sample only: verified sites, no nofollow flag, no first-seen date	API key (one per user, for all their verified sites) or OAuth, on the JSON endpoint. A weekly snapshot diff gives new and lost links. Until the client exists: the community zizzfizzix/mcp-server-bwt, audited and pinned	$0
3	Moz API (Free; Starter Medium during active link building)	Cross-check	A second independent index with DA, PA, Spam Score, Link Intersect (up to 6 targets) and 60-day gained and lost linking domains. Monthly on Free: DA, PA and Spam Score for own site, 3 competitors and shortlisted prospects (≤ 50 rows), against DataForSEO’s rank and spam score and Ahrefs DR. When two of three call a prospect weak, drop it. Starter Medium adds monthly gained and lost domains against DataForSEO’s lists: a loss both report is real	JSON-RPC API (x-moz-token header) or the official MCP (beta, paid plans, account owners) with a Bearer key. Outside the guard, so only Free or Starter Medium, which has no overage. The free tier needs a card at sign-up, which the user enters	$0 on Free. $20 on Starter Medium, only while link building is an active goal
4	Ahrefs domain-rating-free (public API v3 endpoint) (to build: client)	Cross-check	Domain Rating for up to 1,000 targets per POST at 0 units: a third authority opinion. A large rank-order gap between DR, DA and DataForSEO rank flags inflated metrics or an index gap, and that domain gets a live check before it reaches the prospect list. Always worth it, because it is free	REST with a v3 key kept in the keychain. The endpoint docs say keys are free on a free account; Ahrefs’ help page says new keys need Lite, so confirm once. Every DR shown needs “Domain Rating by Ahrefs” with a link next to it	$0
5	page_audit.py --facts on the linking page	Cross-check	Ground truth: whether the link exists now, and its rel and anchor. Run it on every link a vendor reports lost before it goes on the reclamation list, and on every new campaign link. Vendor says lost but the page links: index lag. Vendor says new but the page shows nofollow or sponsored: reclassify. Always worth it	Built in. Check the host with robots_check.py first, and send one request at a time per host. Raw HTML only. Challenges are reported, never bypassed	$0

Big sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	DataForSEO Backlinks API	Main	As Small, for own site + 10 competitors, with one domain_intersection per market (the ka, ru and en competitor sets) and our domain excluded	As Small	Weekly deltas ≈ $3.80, monthly intersections and about 10,000-target scoring ≈ $1–10. ≈ $5–15
2	Bing Webmaster Tools API (to build: Bing client)	Main	As Small. Also the cheapest independent check on both paid indexes for the own site	As Small. Page through GetLinkCounts, then GetUrlLinks per target page, spread across the week. Quotas are [UNVERIFIED], so back off on errors and keep the last snapshot	$0
3	Ahrefs API v3, Standard (to build: REST client)	Cross-check	The deepest second index: link-intersect-refdomains, broken-backlinks, referring-domain and DR history, and new and lost links with history=since:<date>. Compares referring-domain counts and trends per domain with DataForSEO monthly. A gap prospect both find goes first; a loss both report is real. DR is the metric most stakeholders know. The Big stack’s only SEO subscription, and it also feeds the keyword, rank and AI-answer cross-checks. Worth it when links are a KPI or those cross-checks are wanted	REST v3 only: ToS §4.4(j) allows automated use only through the API, and the MCP terms forbid scripts. The client prices calls as max(50, rows × field costs) and checks subscription-info/limits-and-usage (free) and the guard’s ledger before each run. 60 requests a minute, 250 rows a request. Cap the key’s units and turn off pay-as-you-go	$249 (Standard, 800,000 units)
4	page_audit.py --facts on the linking page	Cross-check	As Small. It decides which of two disagreeing indexes is right	As Small, batched in a subagent	$0
5	Ahrefs domain-rating-free (public API v3 endpoint) (to build: client)	Cross-check	As Small: a free third authority opinion beside DataForSEO rank and, when Standard is paid, Ahrefs’ own metrics. It scores the about 10,000 prospect targets at 0 units, which saves Standard’s units for history and link lists. Always worth it, because it is free	As Small, 1,000 targets per POST	$0

Reserve

Tool	Use when …
Moz API (big sites)	Stakeholders report DA or Spam Score, or a disavow discussion needs a second spam opinion. Starter Large $75 (15,000 rows) or Growth Small $125 (50,000). Both bill overage ($5 per 1,000 rows, $25 per 10,000) and the guard can’t see Moz, so add a guard entry first or stay on Starter Medium. In the default free stack the Free tier is a Cross-check
Ahrefs API v3, Lite (small sites)	Competitive link building, or a penalty investigation that needs link history, is the plan’s main goal and $129 fits. REST v3 only, never the MCP in scheduled runs
Semrush backlinks (official MCP)	Semrush is already paid (see the keywords Reserve) and one backlinks_overview per domain a month (40 units) is wanted as a further opinion. Never per-line backlink lists in pipelines (40 units a line). Backlink Audit toxicity is UI-only
Majestic (API plan)	An unnatural-links manual action is open, or a migration link audit is scheduled. Take the API plan ($399.99, or $333.33 a month billed annually) for that month and call GetIndexItemInfo (100 items a call: Trust Flow, Citation Flow, topical Trust Flow). Lite and Pro have no API. No MCP, and the guard doesn’t know it
SE Ranking Backlinks API	SE Ranking already holds the wallet and the guard has its prices. Its daily new and lost endpoint (10 credits a target plus 1 a date point) is a cheap change detector. Per row it costs about $0.0002 against DataForSEO’s $0.000036, and SE Ranking doesn’t say it runs its own crawler
Search Console Links report	A person is present, quarterly or during an unnatural-links manual action: export “Latest links” and “More sample links” (up to 100,000 rows each) into seo-work/. The Search Console API has no links resource
Ahrefs Free (UI)	A person is present and there is no paid Ahrefs plan: a quarterly export of the verified site’s top 1,000 backlinks and its broken backlinks
Common Crawl web graph (paid lists)	A free, citable domain-level view across 50 or more competitor domains is needed, such as a quarterly market study (the Aug–Oct 2026 domain graph has 135.0M nodes and 1.8B edges; the edges file is 7.1 GiB). No anchors, URLs or dates. Common Crawl recommends legal advice before commercial use. In the default free stack it is Main
Moz Pro	A stakeholder works in Moz Pro’s interface. Moz Pro includes no paid API access, so no run reads it

Technical

Can every page that should rank be crawled, rendered and indexed? Is it fast for real users, and is its structured data eligible? What do search and AI bots actually fetch, and did a shipped fix take effect?

Small sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	crawl.py	Main	The core crawl: broken links, links to redirects, duplicates, click depth, orphans against the sitemap, bad canonical targets, hreflang return links and a soft-404 probe, plus page_audit and schema_rules on every page. Its CSV feeds triage.py and the research store. Raw HTML only, one host per crawl, near-duplicates on the first 600 pages	Standard library, no key, any scheduler. It obeys robots.txt, stops when robots.txt returns 5xx or 429, and reports a Cloudflare challenge without bypassing it	$0
2	sitemap_check.py	Main	The URL inventory that sets the class (total_urls). Checks lastmod, file limits, and a sample of listed URLs for non-200, noindex and non-canonical pages	Run --urls-out first, so later steps reuse the list	$0
3	robots_check.py	Main	Google-style robots.txt evaluation, plus the access matrix for 22 search and AI crawlers and Content-Signal lines. --file tests a draft before deploy, so the run blocks a bad change rather than reporting it	It has no --fail-on, so CI wraps it in the documented one-liner	$0
4	page_audit.py	Main	One URL per template: headers, redirect chain, canonical, hreflang (it maps “ge” to “ka”), headings, JSON-LD and challenge detection. In CI, --file --fail-on high blocks a regression before it ships	CI downloads the preview HTML (with x-vercel-protection-bypass where needed) and runs page_audit.py --file --base-url <production URL> --fail-on high. A live run covers headers	$0
5	redirect_check.py	Main	Checks that every old URL in a redirect map 301s or 308s to its new URL in one hop, and that the new URL is a 200, indexable, self-canonical page. This is how a run knows a shipped redirect fix took effect	A CSV map with old and new columns	$0
6	Search Console URL Inspection API (through claude-seo)	Main	Google’s own verdict per URL: indexed or not, Google-chosen against declared canonical, last crawl, robots state, and detected rich results with their issues. 2,000 a day per property covers a small site. Indexed version only, no live test	Measurement owns the quota and schedule; Technical reads the results. claude_seo.py run gsc_inspect.py --site-url <property> --batch urls.txt --json on the sign-in or service account, webmasters.readonly, 600 a minute	$0
7	schema_rules.py	Main	Google’s rich-result requirements as pure data, run on every crawled page and on the rendered DOM. Knows retired features (FAQ, HowTo, sitelinks search box). Re-check it when Google changes its search gallery	Pure function inside the scripts	$0
8	psi.py (PageSpeed Insights API v5)	Main	The Lighthouse lab audit per template: LCP element, biggest opportunities, failed SEO audits. Google plans to drop CrUX field data from this API, so field data comes from rank 9. Small sites often have no CrUX data, which makes lab data the main signal	Set psi_api_key: Google recommends a key for automated queries and publishes no quota. On 429 the run continues without PSI	$0
9	CrUX API and History API (to build: in psi.py)	Main	Field p75 LCP, INP, CLS, FCP and TTFB per origin or URL and form factor, LCP subparts, and up to 40 weekly points. Often origin-only or empty on small sites: record “no field data” and use lab	Free Cloud key, 150 queries a minute per project, which paying can’t raise. History refreshes on Mondays, so the monitor runs on Tuesdays	$0
10	render_diff.py	Main	Raw HTML against the rendered DOM: a title, robots or canonical that changes after JavaScript, and content or links that exist only after rendering. One URL per template settles whether a rendering crawler is needed at all	Rendered DOM from rank 11	$0
11	claude-seo render_page.py (claude_seo.py render)	Main	The unattended renderer for render_diff.py: Playwright Chromium with claude-seo’s SSRF guards, one python3 command, so it works in claude -p and launchd jobs that allow only Bash(python3 *)	Installed on first use (git clone of main, a venv and Chromium, about 150 MB) and pulled before every use. Headless Chrome (--dump-dom, default User-Agent: a Googlebot UA from a non-Google IP looks spoofed) renders only when claude-seo can’t be set up. Own site only	$0
12	log_analyze.py	Main	What Googlebot, Bingbot and AI bots actually fetched: crawl waste, errors served to bots, sitemap URLs never crawled, DNS-verified fake Googlebots. Lower here because shared hosts often give no complete logs	Logs delivered to the run (host export, rsync, an S3 pull). Combined format and .gz only. --verify-dns	$0
13	Chrome DevTools MCP (official)	Main	Diagnoses a poor field INP or LCP with performance traces and insights, throttling and User-Agent emulation. It can save a DOM after interaction. Its Lighthouse audit has no performance category	Local stdio with --headless --isolated --performance-crux=false --usage-statistics=false: by default it sends trace URLs to CrUX and usage statistics to Google. Test it once under launchd	$0
14	IndexNow (to build: submitter)	Main	Finishes the job after a fix ships: changed URLs reach Bing, Yandex, Seznam, Naver and Yep. Google doesn’t use it	Measurement owns the submitter and schedule. Autonomy level 3 and above	$0
15	Bing Webmaster Tools API (GetCrawlStats, GetCrawlIssues, GetUrlInfo, FetchUrl) (to build: Bing client)	Cross-check	Compares Bing’s crawl counts and errors with log_analyze.py’s Bingbot hits and crawl.py’s errors, and FetchUrl’s result with our own fetch, which catches a CDN or WAF serving bots differently. Free, so weekly in the technical monitor, and always when there are no logs	The same Bing client as Measurement	$0
16	DataForSEO On-Page API (guard: page cap)	Cross-check	A second crawler from other IPs, with optional JavaScript or browser rendering and microdata validation. Compares status codes, redirects, canonicals, duplicates (including pages past crawl.py’s first 600) and structured-data errors with crawl.py. Worth it quarterly with the full audit, after a release or migration, or when render_diff.py shows a template depends on JavaScript	Connector, Standard queue. As shipped the guard clamps max_crawl_pages to max_rows (50), so a full check needs its own page cap. Never set robots_txt_merge_mode=override	One 2,000-page crawl a quarter: $0.30 raw, $3.00 with JavaScript, $10.20 with browser rendering. ≈ $0.10–3.40
17	Technical drift (claude_seo.py drift) (through claude-seo)	Cross-check	As in the default stack: each crawl against each URL’s known-good baseline, so a release that drops a canonical, schema block or H1, or adds a noindex, shows up in the next run’s triage	One snapshot per crawl run from crawl.py --csv, the first run only baselines, --rebaseline after verified fixes	$0

Big sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Screaming Frog SEO Spider (licensed), CLI	Main	Crawls the whole site with Chromium rendering, structured-data validation, near-duplicates, custom extraction, crawl comparison, list mode for migrations, and Search Console, PSI and GA4 joined per URL	--headless --crawl … --config … --save-crawl --export-tabs … --project-crawl-comparison true from launchd or cron, on a machine with licence.txt and the accepted EULA. A licence per person who runs crawls. Keep the Ahrefs, Majestic and Moz integrations off (they spend those vendors’ units), and its URL Inspection join off (it spends Measurement’s quota)	£199 a year per user (≈ £17 a month)
2	crawl.py, per section	Main	A raw-HTML crawl per --path-prefix section, up to 2,000 pages each, in parallel subagents. Its output is the only crawl triage.py and the research store can ingest, so it stays Main until a Screaming Frog CSV adapter exists. A section that differs from Screaming Frog flags JavaScript dependence or drift in Screaming Frog’s saved config	One section per subagent	$0
3	sitemap_check.py	Main	As Small, per section and locale. It stops at 200 sitemap files, so on very large sites run it per child index	As Small	$0
4	robots_check.py	Main	As Small. Test facet and parameter rules with --path	As Small	$0
5	log_analyze.py	Main	Crawl budget is a real question at this size, and only logs answer it. CDN logs (usually JSON) need converting first, and no converter is built	As Small, logs pulled on a schedule	$0
6	Search Console URL Inspection API (through claude-seo)	Main	2,000 a day can’t cover the site, so inspect changed URLs, money pages and a rotating sample per template. Whether separate URL-prefix properties per locale folder each get 2,000 a day is [UNVERIFIED]	As Small	$0
7	CrUX API and History API (to build: in psi.py)	Main	Big sites more often have URL-level field data: sample the top URLs per template plus the origin	As Small	$0
8	psi.py	Main	The lab audit per template, which explains poor field numbers	As Small	$0
9	page_audit.py	Main	Each key template in full, and the CI gate that stops a template regression before it reaches thousands of URLs	As Small	$0
10	redirect_check.py	Main	Migration maps and shipped redirect fixes at scale	--workers 8	$0
11	schema_rules.py	Main	As Small	As Small	$0
12	render_diff.py	Main	Decides which sections need Screaming Frog’s rendering, which is far slower	As Small	$0
13	claude-seo render_page.py (claude_seo.py render)	Main	As Small	As Small	$0
14	Chrome DevTools MCP	Main	As Small	As Small	$0
15	IndexNow (to build: submitter)	Main	After each release, changed URLs in batches of up to 10,000	As Small	$0
16	DataForSEO On-Page API (guard: page cap)	Cross-check	Compares status codes, redirects, canonicals, duplicates and microdata errors with Screaming Frog on sampled sections (5,000–20,000 pages). It catches geo-, IP- or User-Agent-dependent serving and WAF blocks that a crawl from one machine misses. Worth it after releases and migrations, and when Screaming Frog and crawl.py disagree	As Small	Per 1,000 pages: $0.15 raw, $1.50 with JavaScript, $5.10 with browser rendering. A monthly 20,000-page sample ≈ $3, $30 or $102
17	Bing Webmaster Tools crawl API (to build: Bing client)	Cross-check	As Small, and against Screaming Frog’s error lists	As Small	$0
18	Technical drift (claude_seo.py drift) (through claude-seo)	Cross-check	As Small, on each section crawl’s CSV	As Small, --workers 8	$0

Reserve

Tool	Use when …
Screaming Frog, licensed (small sites)	Every template depends on JavaScript and crawls recur weekly, or near-duplicates must be checked past 600 pages. £199 a year
Screaming Frog free mode	A person wants a GUI cross-check of crawl.py on 500 URLs or fewer. No scheduling, saving, rendering or integrations, so it can’t be a pipeline step
Screaming Frog MCP (licensed)	A person explores a saved crawl in a watched session. It needs the app running or Claude Desktop, and only the licensed user may use it
Screaming Frog Log File Analyser	Logs arrive in formats log_analyze.py can’t read (W3C/IIS, Amazon ELB, HAProxy, JSON, CSV) and a person is present. GUI only, £99 a year
Sitebulb (Desktop, Cloud, read-only MCP)	The team already licenses it with scheduled audits, and the run reads finished audits through its MCP (50 sample URLs per issue). The MCP can’t start or schedule a crawl
Lumar (GraphQL API, MCP add-on)	Hundreds of thousands to millions of URLs, crawls must run in the cloud with no machine awake, or the organisation already pays for it. Its API can start crawls (runCrawlForProject) with service-account keys. Quote-only
Botify	1M+ URLs that need crawl, logs and Search Console in one platform, or it is already licensed. Its REST API can create and launch an analysis. Enterprise, quote-only, no official MCP found
SE Ranking Website Audit API	The guard has SE Ranking’s prices. It then becomes a cloud-crawler cross-check like On-Page: a standard audit costs 2 credits a page (≈ $0.0004 at the wallet rate), and status and report reads are free
Built-in browser or Claude in Chrome	A person is present and a page needs their session, a challenge they pass themselves, or a Search Console or Bing export with no API. Neither runs in claude -p or launchd jobs
Google Rich Results Test	A new template or markup type needs Google’s verdict before it is indexed. A person runs it. No API
Schema Markup Validator	Markup uses schema.org types beyond Google’s rich results. No API. An unattended version would load the downloadable schema.org vocabulary (V30.1, CC BY-SA 3.0) into schema_rules.py; not built
CrUX on BigQuery	Core Web Vitals must be split by country (for example country_ge) or compared across many origins monthly. Needs a Cloud project with a card on file
Search Console Crawl Stats and Pages reports (UI)	No logs exist and a person exports Googlebot’s own crawl breakdown. No API; Crawl Stats exists for root-level properties only
Ahrefs Site Audit	The site already pays Ahrefs Lite or higher and has a scheduled audit project. Read it through REST v3 (issues cost 50 units a request). The API can’t start a crawl
Semrush Site Audit	The site already pays Semrush and stakeholders want its dashboard. Read the existing project through the MCP (≈ 400 units plus 100 per non-zero issue). Never call “snapshot info” (10,000 units)
Cloudflare AI Crawl Control	The site is on Cloudflare: check its AI-bot policies, which can put rules at the top of robots.txt. Dashboard only; CDN changes are gated at every level
Google Indexing API	The site has JobPosting or BroadcastEvent pages. It is not a general way to get pages indexed
Wayback Machine CDX API	A site move has no list of old URLs

Rank and AI-answer visibility

Where does the site rank per market, language and device, in Google and Bing and in AI answers (Google AI Overviews and AI Mode, ChatGPT, Claude, Perplexity, Copilot)? Which competitors and cited domains sit around it, and did any of that change enough to act on?

Small sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Search Console API, Search Analytics (through claude-seo)	Main	First-party clicks, impressions and average position by query, page, country, device and date. AI Overview and AI Mode appearances count inside Web. It feeds gsc_analyze.py and the per-country CTR curve behind visibility and share of voice. Show it beside snapshot rank, never averaged with it. Don’t compare impressions or position across 2025-05-13 to 2026-04-27	The same claude_seo.py gsc-export as Keywords. Each run re-pulls only the last ~3 days. 1,200 queries a minute per site	$0
2	DataForSEO Google Organic SERP API (AI Overview on the money tier)	Main	The only compliant source of fixed-condition rank snapshots and competitor positions: Google has no SERP API, and seo-suite must not query google.com. One task gives rank, URL, SERP features, AI Overview presence and its reference domains. Georgia 2268 (Tbilisi 1007469, Batumi 1007466; the US state 21143 is a trap), with ka, ru and en	Connector, Standard queue, up to 100 tasks a POST, free tasks_ready. Validate locale pairs against DataForSEO’s lists at the start and fail loudly on an unsupported one. Snapshots go to the research store; the rank-snapshot store is to build	Per pair, at the B6 sizing (50 money keywords daily at depth 10, 450 weekly at depth 20): $3.22, plus $0.90 for AI Overview, plus ≈ $0.65 for daily runs during 2 weeks of rollouts a year. 2 pairs ≈ $9.50
3	ai_probe.py: OpenAI, Anthropic and Perplexity APIs	Main	The only route to automated ChatGPT, Claude and Perplexity answers within the vendors’ terms. Rows are labelled “API stand-in”: OpenAI says its search API uses the same models and tool as ChatGPT Search, and Perplexity says API results differ from its UI. Locale through user_location (GE, Tbilisi), untested. Prompts written natively in ka and ru, and a brand alias table covers Latin, Mkhedruli and Cyrillic spellings. Rows go to ai_visibility.py for Wilson intervals	Three API keys. Model and tool version pinned per row, 3 runs per prompt. Routine probes on a pinned model with one search (for Claude claude-sonnet-5-5 with max_uses=1; Haiku is not allowed). Outside the guard’s hook, so the collector reads other_monthly_usd, per_run_cap_pct and the guard’s ledger, records its own spend there and stops itself (budget settings). Perplexity’s API Terms §5.1 make its performance confidential information, so named cross-vendor metrics that include Perplexity get legal review before any client report	20 prompts × 2 languages × 3 engines × 3 runs = 360 probes: $2.70 in search fees plus tokens. ≈ $3–6 [estimate: about 10,000 tokens in and 1,000 out per probe]
4	DataForSEO Google AI Mode API (guard: price row)	Main	The only automatable AI Mode answers and cited references per keyword and locale. Its 43 languages include ru and en but not ka, so reports say GE-ka AI Mode is not measured	Same connector, ai_mode/task_post in the Standard queue. The guard has no AI Mode row and prices it at half ($0.0006 against $0.0012)	50 money keywords × GE-ru weekly ≈ $0.26
5	Bing Webmaster Tools API (GetQueryStats, GetPageQueryStats) (to build: Bing client)	Main	Free first-party Bing positions, clicks and impressions per query and page. Bing grounds Copilot. The methods take no country, device or date range, so the data is labelled “Bing, all markets”	Weekly pull into its own table. gsc_analyze.py already maps Bing’s columns	$0
6	Google Search Status Dashboard feed	Main	Google’s own start and end dates for updates. The pipeline moves the money tier to daily during a rollout, holds alerts, and compares only a full week after the rollout ends	GET https://status.search.google.com/incidents.json once a run. No key; robots.txt allows it	$0
7	SerpApi, Free plan (to build: collector)	Cross-check	A second SERP collector with its own parser and infrastructure. Compares rank_absolute for the same keyword, locale and device on the same day with DataForSEO, only for keywords about to alert (entering or leaving the top 3 or top 10, or a move of 3+ positions held for 2 snapshots). Alert when both agree within 2 positions. Otherwise record “collector disagreement” and re-check on the next run. Worth it whenever an alert reaches a client	REST with api_key from the environment, gl=ge, hl=ka or ru. Outside the guard: the free plan has no overage, and the collector counts its own calls. The Legal Shield covers only Production ($150) and higher	$0 within 250 searches a month (50 an hour)
8	GA4 AI-referral sessions (analytics-mcp)	Cross-check	Real sessions from chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com and gemini.google.com landing on cited pages. Compares the direction of ai_probe’s citation rate per engine with referral sessions to the pages that engine cites. Citations rising while referrals stay flat mean the prompts miss real demand, or citations earn no clicks. Direction only, never levels. Free, so in every monthly report	Read-only, service-account impersonation, analytics.readonly	$0

Big sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Search Console API (through claude-seo)	Main	As Small, with one request a day per country filter. The CTR curve is fitted per country and device and refitted quarterly	As Small. One property’s requests at a time; on a quota error, retry once after 15 minutes, then fail loudly	$0
2	Search Console bulk export to BigQuery	Main	Removes the 50,000-row cap and flags anonymised queries. Average position = SUM(sum_top_position) / SUM(impressions) + 1	As in Keywords; shared and counted once	As in Keywords
3	DataForSEO Google Organic SERP API	Main	Snapshot rank, SERP features, AI Overview and competitor domains for the whole locale matrix, desktop and mobile. Share of voice and AI Overview citation share for us and every competitor come from the same data. Keyword sets are versioned	As Small. Tag each task with project, locale, device and keyword-set version	200 keywords × 6 pairs × 2 devices weekly at depth 20 with AI Overview ≈ $19, a daily money tier (50 × 6 × 2) ≈ $11, a monthly depth-100 run ≈ $14.40 (the guard caps depth at 50 until limits are per endpoint), and a rollout allowance ≈ $5. ≈ $50
4	ai_probe.py	Main	As Small, plus share of voice per engine, language and prompt class, a citation-source gap, and accuracy grading against a facts sheet. “Wrong” claims go to human review before any client report	As Small	40 prompts × 3 languages × 3 engines × 3 runs = 1,080 probes: $8.10 in search fees plus tokens. ≈ $10–18 [estimate]
5	DataForSEO Google AI Mode API (guard: price row)	Main	As Small, for every non-ka pair (ru, en, he and de are supported)	As Small	200 keywords × 5 pairs weekly ≈ $5
6	Bing Webmaster Tools API (to build: Bing client)	Main	As Small	As Small	$0
7	Google Search Status Dashboard feed	Main	As Small. It matters more where a rollout moves thousands of tracked pairs at once	As Small	$0
8	Ahrefs Rank Tracker via REST v3 (to build: REST client)	Cross-check	Positions collected independently for the same keywords and countries (ge included). Compares each keyword × country × device weekly with DataForSEO. Pairs that differ by more than 3 positions, or sit on opposite sides of the top-10 line, hold their alerts and are re-checked next week. A persistent gap on one locale points to a location or language mismatch in one tool’s setup. Worth it whenever Ahrefs is in the stack, because the reads are free	REST v3 only: management/projects and management/project-keywords set it up at 0 units, and rank-tracker/overview reads it. Standard tracks 2,000 keywords, so keep to the money and mid tiers	$0 extra on the Ahrefs Standard plan priced under Backlinks. Dropped if that plan is dropped
9	Ahrefs Brand Radar custom prompts via REST v3 (to build: REST client)	Cross-check	The same frozen prompts, run by another collector. Ahrefs enters prompts in the web version of each chatbot, except Claude, which it tracks through the API. So under the consumer-app rule only its Claude data and its Google AI Overview and AI Mode data are stored; its ChatGPT, Perplexity, Gemini, Copilot and Grok data is not. Compares Claude mention and citation rates with ai_probe, and its AI Overview and AI Mode cited domains with DataForSEO’s references, both with Wilson intervals. Worth it when AI visibility is a KPI reported to stakeholders	REST v3: prompts added free with management/brand-radar-prompts (countries includes ge). Reads of custom-prompt data cost no units. Each Claude check uses 8 checks. Checks run monthly	Custom prompts are included on Standard (300 checks). A Basic pack is $50 for 2,500 checks. One monthly cycle of the 120 prompts (40 × 3 languages) at 10 checks each (8 for Claude, 1 each for AI Overview and AI Mode) is about 1,200 checks, which fits the 300 plus a Basic pack. Weekly checks would be about 5,200 a month and need the $100 pack (7,000). Packs bill overage at month end ($0.020 a check on Basic), so the REST client counts each cycle’s checks against the included total and stops before it. ≈ $0–50 at monthly checks
10	DataForSEO Bing Organic SERP API	Cross-check	Per-locale Bing ranks under fixed conditions, which Bing Webmaster Tools can’t give. DataForSEO’s Bing languages lack Georgian, so Georgia is GE-ru and GE-en only. Compares direction and order of magnitude with Bing Webmaster’s all-market averages. Worth it when Bing has 10% or more of the site’s search clicks, or AI visibility is a KPI	Same connector and guard, Standard queue	$0.0006 a SERP. 200 keywords × 5 pairs weekly at depth 20 ≈ $5
11	GA4 AI-referral sessions	Cross-check	As Small, per market, against ai_probe and Brand Radar	As Small	$0

Reserve

Tool	Use when …
Search Console Search generative AI report (export)	A person exports it monthly into seo-work/. It gives Google’s own AI Overview and AI Mode impressions per page and country, the first-party check on vendor-sourced Google AI data. No API. Runs read the newest file and flag it stale after 35 days
Bing Webmaster Tools AI Performance report (export)	A person exports it monthly: Copilot and Bing AI citations, grounding queries, Citation Share, Intents and Topics. It is the only compliant Copilot signal. No API found. Same staleness rule
DataForSEO LLM Mentions	A covered market is a target. Its chat_gpt data is US-English only and its collection method isn’t documented, so only its google platform data qualifies under the consumer-app rule. $0.10 a task plus $0.001 a row
DataForSEO LLM Responses	The owner wants one DataForSEO credential instead of three API keys, so it becomes ai_probe’s backend. Not a cross-check: it calls the same vendor APIs. Prompts are capped at 500 characters and Standard results can take 72 hours. Never with Gemini web search
DataForSEO LLM Scraper	The owner explicitly approves vendor-scraped app answers for a one-off parity check of a few prompts, labelled “vendor-scraped” and kept out of trend lines. It scrapes chatgpt.com and gemini.google.com. The B8 pipeline in seo-automation still samples with it and should move to ai_probe
Ahrefs Rank Tracker and Brand Radar (small sites)	Ahrefs Lite or higher is already paid and the REST client exists
SerpApi, Free plan (big sites) (to build: collector)	The site has no Ahrefs plan, so Rank Tracker isn’t there as the second collector. Then use it as in the small list: compare positions with DataForSEO only for keywords about to alert, within 250 searches a month (50 an hour). In the default free stack it is Main
Search Console bulk export (small sites)	Search Console API pulls hit 50,000 rows a day per search type. In the default free stack it is Main
DataForSEO Bing SERP (small sites)	Bing has 10% or more of the site’s search clicks
Semrush Position Tracking	The owner already runs a campaign that stakeholders read in Semrush. Read its overview (100 units a request) monthly. Positions cost 100 units a keyword, so never as a pipeline
Semrush AI Visibility Toolkit	The owner already pays for it ($99 a month per user) and wants a dashboard for people. Its data isn’t in the API or MCP
SE Ranking Rank Tracker and AI Results Tracker	The site is big, has no Ahrefs plan, and wants a second position source: Core $129 or Growth $279 include API access. Its AI tracker data qualifies only once SE Ranking documents an API-based collection method
Otterly.AI (Standard $189: REST and MCP)	The site is big, has no Ahrefs plan, and Otterly documents API-based collection for an engine; only those engines’ data is stored. The pages checked don’t document its method, so none of its engine data qualifies today. Lite has no API
Peec AI	A client already uses it. Prices, which plans include the API and MCP, and its collection method are not published
Profound	An enterprise client contracts it. API on Enterprise only; the trial has none
Gemini API without grounding (“gemini-parametric”)	GA4 shows gemini.google.com at 10% or more of AI-referral sessions for two months [heuristic] and a no-search signal is still wanted. Report it separately as what the model knows without search
Gemini API Grounding with Google Search	Google’s Gemini API Additional Terms change to allow caching, analysing and collecting Links from Grounded Results
Grounding with Bing Search	Microsoft’s terms change to allow storing output and using it to evaluate models
Ahrefs MCP (interactive)	A person is in the session and wants ad-hoc answers from Ahrefs Rank Tracker, Brand Radar or GSC data (Rank Tracker and GSC tools cost 0 units). Never scheduled
Manual checks in the AI apps, by the user	Before the first client report, and when an engine’s API citation rate moves outside its previous 95% interval two months running [heuristic]. The user checks 5 prompts by hand and logs them as surface “app”. Claude never drives the apps. Perplexity’s ToS 5.2(i) also covers systematic manual monitoring, and Copilot is for personal use only
Local Falcon	Map-pack rank for local businesses: see Measurement, local and indexing (DataForSEO Google Maps is Main there)
SISTRIX Visibility Index	A target market SISTRIX covers (for example DE or GB) and stakeholders report in its index. No Georgia or Israel. API not checked
Semrush Sensor	In a watched session, a drop matches no Search Status incident yet. A free page with no API; Semrush’s ToS bars scripted extraction

Measurement, local and indexing

What did our own site get from Google and Bing (clicks, impressions, positions, indexing state, conversions)? Did the engines hear about our changes? How visible and accurate are our local listings and map-pack ranks? Local tools apply only to businesses with physical locations or service areas, and Merchant Center only to stores. Otherwise they are skipped, not replaced.

Small sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Search Console API: Search Analytics, URL Inspection, Sitemaps (through claude-seo)	Main	Google’s own clicks, impressions and positions for 16 months, and its indexing verdict per URL. A small site fits the 2,000 inspections a day: every URL weekly, changed URLs daily. sitemaps.submit lets a level-4 run submit a sitemap itself. Owns the URL Inspection quota for every category	Service account as a property user: webmasters.readonly, plus webmasters only for sitemaps.submit at level 4 (claude_seo.py submit-sitemap). Never the Indexing API scope. One day per request with dataState=final. Google publishes no Search Console MCP. Without a sign-in or service account: exports. Never script the Search Console UI	$0
2	Google Analytics 4 (official analytics-mcp)	Main	What organic visitors do and whether they convert, by landing page, and GBP traffic through utm_campaign=gbp. GA4 organic sessions against GSC clicks per landing page flag a removed tag or a consent change before a report blames rankings	Local stdio (Experimental), service-account impersonation, analytics.readonly, never --allow-write. A standard property allows 200,000 tokens a day; send returnPropertyQuota	$0
3	Bing Webmaster Tools API (to build: Bing client)	Main	Bing’s own query and page stats and crawl issues, for all markets at once (URL submission is Reserve: IndexNow submits). gsc_analyze.py already reads Bing’s columns	API key or OAuth on the JSON endpoint (SOAP and POX were scheduled to retire on 2026-08-31). The key travels in the query string, so never log request URLs	$0
4	IndexNow (to build: submitter)	Main	The run finishes “tell the engines” itself: changed and removed URLs reach Bing, Yandex (for ru audiences), Seznam, Naver and Yep, which share submissions. Google’s side is a clean sitemap plus sitemaps.submit	Key file /{key}.txt added by a level-2 pull request. POST up to 10,000 URLs at level 3 and above, only changed URLs. 422 means a host mismatch; 429 means back off	$0
5	Google Business Profile APIs (local businesses only) (to build: client)	Main	Daily Maps and Search impressions, calls, directions, website clicks and bookings, monthly search keywords, and own reviews. The only official way for a run to apply an approved profile change, which stays gated per run	Google must approve access first: a profile verified and active for 60+ days, a website, and the application form. 0 queries a minute until then, 300 after. OAuth business.manage as owner or manager, read methods only unless an edit was approved for that run. De-duplicate reviews (a pagination known issue since 2026-06-17)	$0
6	Merchant API v1 (stores only)	Main	Product statuses, disapprovals, account issues and performance: the input to seo-ecommerce’s feed, page and markup check. Content API for Shopping was sunset on 2026-08-18	Service account, scope content. Read tools only, through the alpha Merchant API MCP (https://merchantapi.googleapis.com/mcp) or plain REST as the fallback. Exclude create_data_source and fetch_data_source. Edits are always gated	$0
7	DataForSEO Google Maps and Local Finder, geo-grid (local businesses only) (guard: price row)	Main	The cheapest unattended map-pack rank: one Maps SERP per grid point (location_coordinate “lat,lng,zoom”), stored raw, so our share of local voice and the competitor list come from one call. Georgian city codes, and ka, ru and en. [vendor data]: we never query google.com/maps	Connector, Standard queue, fixed zoom and depth per keyword set. Avoid calculate_rectangles (2×) and operators (5×). The guard bills Maps per 10 results while DataForSEO bills per 100, so it over-reserves	$0.0006 a point. 1 location × 10 keywords × a 5×5 grid ≈ $0.15 monthly or ≈ $0.65 weekly
8	Google Search Status Dashboard feed	Main	Dates of Google updates and incidents, for drop diagnosis	As in Rank; shared and counted once	$0
9	Microsoft Clarity (Data Export API, official MCP)	Cross-check	A second tag that counts sessions its own way, with bot sessions separated, plus quickbacks that show an intent mismatch. Compares organic sessions and engagement per landing page with GA4 and GSC clicks. A gap above about 20% [heuristic] points to consent, tagging or bot-filtering differences. Worth it when Clarity is installed, or when GA4 and GSC disagree	Token created by the project admin. 10 requests a project a day, the last 1–3 days, 3 dimensions and 1,000 rows, so one fixed daily job. Adding the tag is a site change (level 2 and above)	$0

Big sites

#	Tool	Role	Why	How a run drives it	Monthly cost
1	Search Console API (through claude-seo)	Main	Fresh days by country, URL Inspection of changed URLs plus a fixed sample per template, and sitemaps.submit at level 4. Comparing the API’s daily totals with the export’s site table shows how much the API truncates	As Small. One property at a time; retry once after 15 minutes, then fail loudly	$0
2	Search Console bulk export to BigQuery	Main	Full query × URL tables for cannibalisation and language splits	As in Keywords. Set partition expiration (14 days minimum) or data accumulates forever. Shared and counted once	As in Keywords
3	Google Analytics 4	Main	By landing page, section and market	As Small. Pull fixed weekly reports rather than many ad-hoc ones (360 properties get 10× the token quota)	$0
4	Bing Webmaster Tools API (to build: Bing client)	Main	As Small, weekly	As Small	$0
5	IndexNow (to build: submitter)	Main	Many URLs per deploy, with no account and no quota beyond 429 throttling	As Small, batches of up to 10,000 per deploy	$0
6	Google Business Profile APIs (local businesses only) (to build: client)	Main	Per-location data across every location, and approved fixes in bulk, gated per run. Location groups map to the site’s location pages	As Small. 300 queries a minute covers hundreds of locations daily	$0
7	Merchant API v1 (stores only)	Main	As Small, for big catalogues	As Small. Cap calls per run: agents can drain the quota it shares with the API	$0
8	DataForSEO Google Maps and Local Finder (local businesses only) (guard: price row)	Main	Every location × money keyword	As Small, with versioned keyword sets	20 locations × 10 keywords × a 7×7 grid = 9,800 points ≈ $5.90 a scan: ≈ $6 monthly or ≈ $25 weekly
9	Google Search Status Dashboard feed	Main	As Small. Drops on big sites are often partial	As Small	$0
10	Local Falcon (API v2, hosted MCP; local businesses only)	Cross-check	Another geo-grid scanner, with its own share-of-local-voice metric, stored trends, and Apple Maps scans that no other tool here measures. Compares our rank and top-3 presence point by point each month with the DataForSEO grid on money keywords. Agreement at 80% or more of points [heuristic] confirms a change before profile work starts. Disagreement means checking coordinates, zoom and language first. Worth it from more than 10 locations, or when one swing of 3+ positions would trigger work	POST /v2/run-scan with api_key, place_id and a grid of 3–21. Reading reports costs no credits. Google and Apple platforms only: its AI-engine scans fall under the consumer-app rule. Outside the guard, so rely on the plan’s credit cap and buy no pay-as-you-go credits	Basic $49.99 (15,150 credits) or Pro $99.99 (31,250), at 1 credit a grid point
11	Microsoft Clarity	Cross-check	As Small, section by section. 1,000 rows won’t cover every URL	As Small, with rotating section filters	$0

Reserve

Tool	Use when …
Yext (Knowledge Graph API, hosted MCP)	More than 10 locations in markets that Yext confirms on a demo (its Georgian publisher coverage is [UNVERIFIED]), and the owner pre-approves listing pushes per run. It publishes one record to 200+ publishers. Quote-only; third parties report $199–999 per location a year [UNVERIFIED]. Pushing to Google through it is a Business Profile edit, which is always gated. Its MCP reportedly has write access on by default [UNVERIFIED]: turn that off or use a read-only key
Local Falcon (small sites)	A map-pack swing is disputed, or Apple Maps rank matters. Starter $24.99 a month
Search Console bulk export (small sites)	Search Console API pulls hit 50,000 rows a day per search type. In the default free stack it is Main
Looker Studio (Google’s pages now title it “Data Studio”), and Pro	A person wants a live dashboard; the run writes the data it reads. Its API manages assets only and can’t read report data
Apple Business (replaced Business Connect on 2026-04-14)	The owner exports Place Card insights, or an Apple-approved partner the owner delegated to reads the Insights API. Otherwise a dashboard
Bing Places for Business	Bing and Copilot listings need correcting, and the owner applies the corrections. No public API is documented
BrightLocal	The business is in the USA, UK, Canada or Australia and needs a citation audit or builds, or a third geo-grid opinion. No Georgia. Track from $41 a month for one location
Semrush Listing Management and Local	The locations are in a country Semrush supports (Georgia is not one) and the account already pays Semrush. Local Pro is $60 per location a month, with an API
Conductor (Data API, MCP)	The organisation already licenses Conductor Intelligence: read its tracked rankings (5,000 calls a day per key) instead of buying the same set
Bing URL Submission API (SubmitUrlBatch)	A URL must reach Bing before an IndexNow key file can be deployed. 500 URLs a batch; Bing sets each site’s daily quota
Google Indexing API	JobPosting or BroadcastEvent pages only (200 a day)
Community MCP (zizzfizzix/mcp-server-bwt)	A run needs Bing Webmaster API tools before the plugin’s own Bing client exists. Audit and pin a version first. Search Console runs through claude-seo
Ahrefs GSC endpoints	A watched session with Ahrefs connected and no Google credential yet. 0 units except gsc-anonymous-queries. The same data as Search Console, so no cross-check value
Ahrefs Web Analytics	The site already uses it instead of GA4. Its API v3 web-analytics endpoints cost 0 units, but need a v3 key and the REST client (to build)
GA4 BigQuery export	GA4 reports hit thresholds or quotas, or event-level joins with the Search Console export are needed. Daily export up to 1M events on standard properties; streaming export $0.05 a GB
Semrush Sensor	As in Rank
Google Places API	Only to resolve a location’s Place ID. Its terms forbid storing names, addresses or reviews
DataForSEO Business Data (Google My Business Info, Reviews)	Comparing map-pack competitors’ categories, ratings and review velocity, which the Business Profile API can’t give. About $0.0015 a profile [UNVERIFIED]. Store counts, not review text
Search Console generative AI report, Bing AI Performance report	See the Rank Reserve

Credit-guard limits

What scripts/credit_guard.py does today, and what that means for the paid lists:

• Unpriced vendors share one small allowance. SE Ranking, AlsoAsked, Similarweb, Otterly and Keywords Everywhere calls, and Lumar crawls, share other_monthly_usd, and each call is counted at ask_above_pct of it. At the defaults (25% a run, 5% a call) that is 5 calls a run and 20 a month for the whole bucket, whatever the budget, and 0-credit calls count too. Raising other_monthly_usd doesn’t change the count, and lowering ask_above_pct is global. A per-vendor bucket and an SE Ranking price table would fix this; until then, plan around it.
• One row limit for everything. max_rows (default 50) also clamps DataForSEO depth, limit and max_crawl_pages, and adds them when missing. So On-Page stops at 50 pages, a 1,000-row Labs pull becomes 20 tasks, and depth 100 is capped at 50. Raising max_rows loosens Semrush and Ahrefs too. Per-vendor and per-endpoint limits (Labs and Backlinks 1,000, SERP depth 100, On-Page up to the site’s URL count) would fix this.
• Missing price rows. AI Mode is priced as a generic SERP, half its $0.0012. Maps is billed per 10 results, while DataForSEO bills per 100. Clickstream, LLM Responses, LLM Scraper and Business Data have no price, so each call is counted at 5% of the DataForSEO budget.
• Tools the hook never sees. Moz, Majestic, Local Falcon, SerpApi, OpenAI, Anthropic, Perplexity, and Ahrefs REST calls. Adding a vendor is a code change; until then they follow the no-overage rule. ai_probe.py and rank_snapshot.py read and record the guard’s ledger themselves.

Costs in the lists are at the intended call shapes. Budget settings for the paid stacks (/plugin configure seo-suite@seo-suite). The default free stack needs none, because it makes no paid calls:

Option	Small	Big
dataforseo_monthly_usd	25	270
other_monthly_usd	12, after the SE Ranking change, for SE Ranking (≈ $0.60) and ai_probe (≈ $6)	35, after the SE Ranking change, for SE Ranking (≈ $3) and ai_probe (≈ $18)
ahrefs_monthly_units	0	Up to 320,000 (40% of Standard), read by the REST client

• ai_probe’s cap. The hook never sees OpenAI, Anthropic or Perplexity calls, so ai_probe.py reads other_monthly_usd, per_run_cap_pct and the guard’s ledger before each batch of probes, records its own spend in the ledger, and stops at the monthly budget or the per-run cap. It counts against other_monthly_usd with SE Ranking. The budgets leave room above the two’s expected spend (about $6.60 small and $21 big), so a re-run still fits. Run the probes weekly, so each run stays under the per-run cap (25%: $3 small, $8.75 big). On big sites, run the quarterly SE Ranking pulls (about $8) in a run without probes.
• Why 270 for big. The big list’s DataForSEO spend reaches about $253 in an average month: Main $105–145, plus the cross-checks On-Page up to $102, Bing SERP about $5 and Clickstream about $0.75. Real months are uneven, because the Labs pull ($22.20) and the Autocomplete expansion (about $35) are quarterly. Run those two in different months of the quarter. The heaviest month is then about $265: recurring Main with the weekly Maps grid (about $123), the Autocomplete expansion, and the cross-checks with browser-rendered On-Page. If a month still runs short, drop cross-checks in this order: browser-rendered On-Page (fall back to JavaScript mode, then raw), DataForSEO Bing SERP, Clickstream. Main calls keep running. With JavaScript-mode On-Page ($30) instead of browser rendering, 200 is enough.
• The per-call threshold. The guard denies any priced call estimated above ask_above_pct (default 5%) of the monthly bucket and defers it to the end-of-run questions, which need a person. At these budgets that is $1.25 a call for small sites and $13.50 for big ones, under a per-run cap (25%) of $6.25 and $67.50. The guard prices an On-Page task_post as one call: the per-page price ($0.00015 raw, $0.0015 with JavaScript, $0.0051 with browser rendering) × max_crawl_pages, summed over the tasks in the POST. So a run splits On-Page into separate task_post calls of one task each, every one under the threshold: about 800 JavaScript or 245 browser-rendered pages a call for small sites, and about 9,000 JavaScript or 2,600 browser-rendered pages for big ones. Raw crawls fit easily (about 8,300 and 90,000 pages). Each task covers one section: respect_sitemap and crawl_sitemap_only set to true, and custom_sitemap set to that section’s child sitemap. Spread the calls over runs that make no other large DataForSEO calls, to stay under the per-run cap: a small browser-rendered crawl ($10.20) takes two runs, and the big 20,000-page browser-rendered sample ($102) takes two runs of about $51. The other way is raising ask_above_pct, which applies to every vendor. All of this needs the page-cap change first, because as shipped the guard clamps max_crawl_pages to 50.

Semrush

Justified when you need competitors’ keywords and estimated traffic, country-level keyword volume and difficulty at scale, PPC/ad intelligence, or market and traffic estimates. These come from Semrush’s ranking database. The free path cannot replicate them.

Shared-access (“group buy”) proxies. A Semrush session on a domain other than semrush.com (for example 1.semrush.com.in) is a third-party reseller. Semrush’s terms may forbid shared accounts, the operator sees every query, data can be cached or partial, and sessions expire without warning. Use it read-only, in the user’s own browser, never type credentials into it, tag its numbers [vendor data: proxy] with the database and date, and log what was queried, since the credit guard never sees browser use.

Stack: Reserve (small, big). Use when the owner already pays Starter or Pro+ and a scheduled run has kept its sign-in working; then it is a third Georgian keyword database and a backlinks opinion. See the Main stack. Free plan (10 reports a day): Reserve in every list, use when a person wants a quick look in the interface; it has no API or MCP access, so no run reads it.

Plans (monthly / annual per month):

• Free: 10 reports/day.
• SEO: $139 / $117.33. 5 sites, 500 tracked keywords. No MCP or API access is listed, so a run can’t use it.
• Starter: $199 / $165.17. 50 AI prompts. MCP access: the cheapest plan that lists it.
• Pro+: $299 / $248.17. 15 sites, 1,500 keywords, history, cannibalisation.
• Advanced and Enterprise: custom.
• Extra seats from $45/mo.
• Rows per report are not stated on the vendor page [UNVERIFIED: Pro 10k, Guru 30k, Business 50k].

API:

• Needs the Business-equivalent tier (Advanced or Enterprise) plus unit packages (2M–20M units, prices not public; unused units expire at renewal).
• Rate limit 10 requests/s. v3 keyword and backlink endpoints are deprecated; use v4.
• Unit costs per returned line (historical = 5×):

Call	Units
domain_organic	10 per line
phrase_this / phrase_these (batch ≤100) / phrase_organic	10 per line
phrase_related / phrase_questions	40 per line
phrase_kdi (difficulty)	50 per line
competitors	40 per line
domain vs domain	80 per line
backlinks_overview	40 per request
backlinks / referring domains	40 per line

• Empty results cost 0 units. Test calls do cost units. The countapiunits balance check is free.

MCP:

• Endpoint https://mcp.semrush.com/v2/mcp (OAuth or Authorization: Apikey KEY), read-only.
• Starter, Pro+ and legacy Pro/Guru include 50,000 MCP units per renewal, shared with API pricing.
• Official connectors exist for Claude, ChatGPT and Perplexity.
• Claude Code: claude mcp add --transport http semrush https://mcp.semrush.com/v2/mcp.

Add-ons:

• AI Visibility Toolkit: $99/mo per user. 1 domain, 25 prompts, ChatGPT + Google AI Mode. Reserve: its data isn’t in the API or MCP.
• Traffic & Market: $289/mo per user annually. The Trends API is a separate quarterly subscription. Reserve: use when an enterprise client needs Semrush’s panel data instead of Similarweb.

Cost tactics (see also the playbook):

1. Always set display_limit, and page with display_offset only when needed. Billing is per line.
2. Never pull historical data (display_date) by default (5× cost). Snapshot monthly yourself.
3. Cheapest call first: overview and organic lines (10) before related/questions (40). Run phrase_kdi (50) only on a shortlist.
4. Batch with phrase_these (≤100 keywords). It saves requests and rate limit, not units.
5. One database (country) per call; avoid all-database pulls.
6. Use backlinks_overview (flat 40) before any per-line backlink list.
7. 50k MCP units ≈ 5,000 keyword lines: ask narrow questions with fixed limits, database and filters.
8. Read existing Projects (Position Tracking, Site Audit) through the Projects API or MCP instead of re-running ad-hoc reports.
9. Cache volumes and difficulty ~30 days; Semrush refreshes volumes monthly.

Ahrefs

Justified when you need the largest fresh backlink index (competitor link profiles, link intersect, lost and new links), organic traffic per URL, Content Explorer, or AI mentions across many engines (Brand Radar). For your own site, use Ahrefs Free first.

Stack: REST API v3 on Standard: Cross-check (big, to build) for backlinks, Rank Tracker, Brand Radar and keyword difficulty, and the Big stack’s only SEO subscription. domain-rating-free: Main in the default stack, Cross-check (small, big), to build. Lite for small sites, Site Audit, the GSC endpoints and the MCP: Reserve (the MCP only in sessions someone is watching). Starter ($29, no API units): Reserve, use when a person wants Ahrefs’ interface for occasional checks; no run reads it. Content Explorer, Keyword Clusters and Batch Analysis are features of Standard and higher, so they carry Standard’s labels: Batch Analysis is the API’s batch-analysis endpoint, used inside the REST v3 Cross-check (cost tactic 2). Content Explorer and Keyword Clusters have no API v3 path, so they are Reserve, for a person in the interface when Standard is already paid.

Plans (monthly):

Plan	Price	Tracked keywords	History	API units/mo
Starter	$29	750	1 month	none
Lite	$129	750	6 months	200k (100 rows/request)
Standard	$249	2,000	2 years	800k (250 rows/request)
Advanced	$449	5,000	5 years	2M (500 rows/request)
Enterprise	$1,499 (annual only)	10,000	unlimited	4M+

• Annual billing saves up to 17%. Seats cost $40, $60 or $80/mo, capped per plan.
• Content Explorer, Keyword Clusters and Batch Analysis start at Standard.

API v3 metering:

• Cost per request = max(50, rows × sum of field costs).
• Most fields cost 1; traffic and org_traffic cost 10; refdomains costs 5. Fields are charged even when used only in where or order_by.
• Cached responses cost 0.
• Free endpoints: Rank Tracker, Management, Public (domain-rating-free) and the GSC endpoints. Web Analytics endpoints are free too, and so are Brand Radar reads that return only custom-prompt data.
• API v3 works directly on Lite and higher (100 rows a request on Lite, 250 on Standard, 500 on Advanced).
• Free test targets: ahrefs.com, yep.com, firehose.com, capped at 100 rows. The response headers show the cost (x-api-units-cost-total, x-api-cache).

MCP:

• Endpoint https://api.ahrefs.com/mcp/mcp (OAuth or MCP key; Lite and above). Units are shared with the API, 50 minimum per call.
• OAuth keys default to a 200k units/month cap.
• Terms forbid scripting the MCP endpoint; use API v3 for programmatic work.
• Claude Code: claude mcp add --transport http ahrefs https://api.ahrefs.com/mcp/mcp.

Brand Radar (AI visibility):

• From $199/mo. Prompt packs: $50 for 2.5k checks, $100 for 7k, $250 for 25k.
• Paid plans include 150, 300 or 600 checks.
• Engines: AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity, Copilot, Grok, Claude.
• Collection: Ahrefs enters prompts in the web version of each chatbot with its default model. Claude is tracked through the API, and each Claude check uses 8 checks. Under the Main stack’s consumer-app rule, only the Claude, AI Overview and AI Mode data is stored.

Cost tactics:

1. Select only the fields you need. Avoid traffic/org_traffic (10×) and refdomains (5×) in select, where and order_by unless they are the point.
2. Beat the 50-unit minimum by consolidating: Batch Analysis takes up to 100 targets per call.
3. Always limit. Estimate first on a free test target and read x-api-units-cost-total.
4. Own-site data is free: Ahrefs GSC endpoints (0 units), Ahrefs Free, verified projects (no row caps).
5. Cap OAuth/MCP keys at a monthly unit limit you chose.
6. Plan history is short on Lite (6 months): snapshot what you’ll need later rather than re-pulling.
7. Spend the Brand Radar checks your plan includes before buying a pack. Packs bill overage at the end of the month ($0.020 a check on Basic, $0.015 on Growth, $0.010 on Scale), so count checks and stop before the included total.

Screaming Frog

Justified when the site is larger than ~2,000 URLs or relies on JavaScript. The plugin’s crawler is raw-HTML only and best up to a few thousand pages. Screaming Frog adds Chromium rendering at scale, custom extraction, crawl comparison, scheduling, and GSC/GA4/PSI joined per URL.

Stack: licensed CLI: Main (big). Small sites: Reserve, use when every template depends on JavaScript and crawls recur weekly. Free mode, MCP and Log File Analyser: Reserve. The default free stack has no licence, so crawl.py per section does the crawling there and Screaming Frog is Reserve.

• Licence: £199/year per user (volume discounts down to £169) [USD ≈ $279 UNVERIFIED]. The free mode covers 500 URLs.
• CLI: headless with a licence.txt key file. Scheduling and saving crawls need a licence. On a machine with no UI, also accept the EULA in spider.config. A licence covers one person.
• MCP: v24.0 (May 2026) added an SEO Spider MCP server for Claude (crawl, analyse, export). List mode is not supported yet. The MCP is a paid licence feature only, and it needs the app running or Claude Desktop.

Cost tactics:

• One licence per person who actually crawls.
• Use the free mode to cross-check crawl.py on small sites.
• Link-metric integrations (Ahrefs, Majestic, Moz) spend those tools’ API units. Enable them only for key URLs.
• Schedule crawls and compare runs automatically instead of recrawling by hand.

Alternatives:

• Sitebulb: Desktop Lite/Pro; Cloud from £95/mo; JS rendering included; MCP on every plan. Reserve: its MCP is read-only and can’t start or schedule a crawl.
• Lumar: quote-only; official MCP at mcp.lumar.io/mcp. Reserve: use when crawls must run in the cloud at very large scale. Its GraphQL API can start crawls.
• Botify and Conductor: enterprise. Reserve: use when the organisation already licenses them.

DataForSEO

Justified when you need raw data at the lowest unit cost, paying per use instead of subscribing:

• Bulk Google Ads volumes.
• SERP snapshots at scale, including AI Overview and AI Mode, for rank tracking or SERP-overlap clustering.
• Backlink data without a subscription.
• LLM mention/response data for AI visibility.

Stack: Main (small, big): the main paid data source in every category of the Main stack. Labels per API are in the table. Reserve (default): it is pay per call, so it comes in when its budget is set or the user asks.

Billing: pay-as-you-go, $50 minimum top-up, and balances never expire. The $100/mo commitment for Backlinks and LLM Mentions was removed on 2026-07-01.

Prices:

API	Price	Stack
SERP (Google organic, per 10 results)	Standard queue $0.0006 · Priority $0.0012 · Live $0.002	Main (small, big): keywords, competitors, rank
SERP extras	AI Overview load 2× · site:-style operators 5× · each PAA click $0.00015	AI Overview: Main on the money tier. PAA: Main. Operators (5×): not used
Keywords Data (Google Ads)	$0.06 per task of up to 1,000 keywords, Standard mode (≈ $60 per 1M keywords); Live $0.09	Main (small, big)
Labs	$0.012 per task + $0.00012 per item	Main (big, markets Labs covers) · Cross-check (small, covered markets). No Georgia. Guard: rows
Backlinks	$0.024 per request + $0.000036 per row	Main (small, big)
On-Page	from $0.00015 per page	Cross-check (small, big). Guard: page cap
LLM Mentions	$0.10 per task + $0.001 per row	Reserve: use when a covered market is a target (google platform data only)
Google Autocomplete (per task)	Standard $0.0006 · Priority $0.0012 · Live $0.002	Main (small, big)
Google AI Mode (per SERP)	Standard $0.0012	Main (small, big). No Georgian. Guard: price row
Google Maps (per SERP of up to 100 results) · Local Finder (per 10 results)	Standard $0.0006 each	Main (small, big) for local businesses. Guard: price row
Bing organic (per 10 results)	Standard $0.0006 · Priority $0.0012 · Live $0.002	Cross-check (big) · Reserve (small)
Clickstream bulk search volume	$0.012 per task + $0.00012 per keyword, Live only	Cross-check (big, covered markets) · Reserve (small). Guard: price
LLM Responses	Standard $0.0002 per task plus a $0.01 advance, with the unused part refunded	Reserve: use when one DataForSEO credential should replace the three ai_probe API keys
LLM Scraper	Standard $0.0012 per task	Reserve: it scrapes consumer AI apps, so only with the owner’s explicit approval for a one-off parity check

MCP:

• Official server github.com/dataforseo/mcp-server-typescript (npm dataforseo-mcp-server), remote https://mcp.dataforseo.com/mcp. The docs list several paths; test which one answers.
• Auth uses your API login and password.
• Trim response fields with FIELD_CONFIG_JSON to save tokens.

Cost tactics:

1. Standard queue, not Live, for anything that isn’t interactive.
2. Fill Keywords Data tasks to 1,000 keywords; you pay per task, not per keyword.
3. Request only the SERP depth you need; every extra 10 results costs another base price.
4. Avoid operator queries (5×).
5. Load AI Overviews only for the keywords where they matter (2×).
6. Cache SERPs for clustering; they don’t need to be fresh to the minute.

Moz

Justified when a client or stakeholder reports in Domain Authority / Page Authority / Spam Score, or you need a second link index to cross-check Ahrefs or Majestic.

Stack: API Free tier: Cross-check (default and small backlinks), with Starter Medium only while link building is active on a small site, because it has no overage. Big sites and keyword metrics: Reserve, use when stakeholders report DA or Spam Score. Moz Pro: Reserve (UI only).

Moz Pro figures are [UNVERIFIED], because moz.com blocked an earlier fetch. The Moz API figures were read from moz.com on 2026-10-24.

• Moz Pro: Starter $49, Standard $99, Medium $179, Large $299 per month (about 20% off annually). Moz Pro includes no paid API access.
• Moz API: sold separately. Free: 50 rows a month, one request every 10 seconds; a card is needed at sign-up. Starter: $20 for 3,000 rows (no overage) or $75 for 15,000 ($5 per extra 1,000). Growth: $125 for 50,000 rows, $250 for 120,000 or $500 for 500,000. Every plan includes URL Metrics, Links, Link Intersect, Linking Root Domains, Anchor Text and Brand Authority. Auth is an x-moz-token header.
• MCP: official and in beta, on paid plans, for account owners: https://api.moz.com/mcp/v1/data, with OAuth or Authorization: Bearer <Moz API key>. On the free tier it is [UNVERIFIED]. The credit guard doesn’t recognise Moz.

Cost tactics:

• Use DA/PA only as a relative comparison. It is not a Google metric.
• Pull metrics for shortlisted prospects only, never whole link lists.

Other paid tools

Tool	Justified for	Price	API / MCP	Stack
Majestic	Trust Flow / Citation Flow, topical link graphs	Lite $49.99 · Pro $99.99 · API $399.99/mo	Full API on the API plan only; no MCP	Reserve: use when an unnatural-links manual action or a migration link audit is open
SE Ranking	Cheaper all-in-one rank tracking + audit + AI prompts	Core $129 · Growth $279/mo · standalone API from a $50 wallet (250,000 credits that never expire)	MCP in every API tier (https://api.seranking.com/mcp)	Data API (to build: guard prices), for competitor keywords and gaps with a Georgia database: Main (small). In the big list it splits by market: Main for the GE pairs and the markets DataForSEO Labs lacks, and Cross-check (big, in markets Labs covers), where Labs is Main. Backlinks, Website Audit, Rank Tracker: Reserve. Reserve (default)
Similarweb	Competitor traffic, sources, audience overlap (estimates)	Quote-only	MCP needs API, Business or Enterprise	Reserve: use when a board-level market-sizing decision is due and an API contract is budgeted
SpyFu	Historical PPC / competitor ads	~$39–$249/mo [UNVERIFIED]	API on Pro + AI ($40 API credit a month) and Team ($100), pay as you go above that	Reserve: use when historical Google Ads keywords and ad copy are needed in its 16 countries (no Georgia or Israel)
Keywords Everywhere	Cheap volume/CPC overlay while browsing; small API jobs	From $84/yr for 100k credits (1 credit = 1 keyword volume)	API and MCP on all plans	Reserve: use for spot checks in the countries its checker offers (no Georgia)
AlsoAsked	People-Also-Ask question trees	From $12/mo billed annually: Basic $15 · Lite $29 · Pro $59 ($47 annual) monthly	API and MCP on Pro only; MCP at https://alsoasked.com/mcp	Reserve: use when a question hub of 100+ pages is planned
AnswerThePublic	Question/preposition idea lists	From $20/mo: $20 · $99 · $199	No API listed	Reserve: use when a person brainstorms seeds by hand
Local Falcon / DataForSEO local	Map-pack geo-grid rankings	Per-scan pricing: Local Falcon from $24.99/mo at 1 credit per grid point; DataForSEO Maps $0.0006 per point	Local Falcon: API and MCP from Starter. DataForSEO: the connector	DataForSEO Maps grid: Main (small, big) for local businesses. Local Falcon: Cross-check (big, local) · Reserve (small)

AI visibility trackers

Justified when you need repeated, automated prompt tracking across engines over time. The free method in seo-ai-search (a frozen query set, 3–5 manual runs per engine, ai_visibility.py with confidence intervals) is enough to start, and to decide whether a tracker is worth paying for. In the Main stack, ai_probe runs the frozen set through the engines’ official APIs instead, and Claude never drives the consumer apps.

Tool	Price	Engines	API / MCP	Stack
Ahrefs Brand Radar	from $199/mo; prompt packs $50–$250	AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity, Copilot, Grok, Claude	API + MCP	Cross-check (big, to build): custom prompts through REST v3, storing only the Claude (API), AI Overview and AI Mode data
Semrush AI Visibility	$99/mo per user (25 prompts); Semrush One bundles 50–200	ChatGPT, Google AI Mode	—	Reserve: use when the owner already pays and wants a dashboard for people
Otterly.AI	Lite $29 (15 prompts) · Standard $189 (100) · Premium $489 (400)	ChatGPT, AI Overviews, Perplexity, Copilot (+AI Mode, Gemini, Claude as add-ons)	API + MCP from Standard	Reserve: use when a big site has no Ahrefs and Otterly documents API-based collection
Profound	Enterprise; 7-day trial	up to 9 engines	Enterprise	Reserve: use when an enterprise client contracts it
Peec AI	prices not shown on vendor page [UNVERIFIED ~€89+]	ChatGPT, Perplexity, Gemini	API, MCP	Reserve: use when a client already uses it
DataForSEO LLM Mentions / Responses	pay per task	ChatGPT, AI Overviews (Mentions); ChatGPT, Claude, Gemini, Perplexity (Responses)	API + MCP	Reserve: Mentions when a covered market is a target; Responses as an alternative ai_probe backend
ai_probe.py	Search fees: OpenAI $10 and Anthropic $10 per 1,000 searches, Perplexity $2.50 per 1,000, plus tokens	ChatGPT, Claude, Perplexity, as API stand-ins	The vendors’ official APIs	Main (small, big). Reserve (default), where GA4 AI-referral sessions are the AI-answer signal

Cost tactics:

• Freeze 25–50 prompts that map to real revenue topics before buying.
• Track weekly, not daily; AI answers vary run to run anyway.
• Prefer the tracker bundled with a tool you already pay for.

Cost-optimisation playbook

1. Free data first, always: GSC, Bing WMT, GA4, PSI/CrUX, Ahrefs Free, Screaming Frog free mode, and this plugin’s scripts. Many questions end here.
2. Shortlist before you pay. Generate ideas free (keyword_ideas.py, GSC, PAA). Pay for volume/difficulty only on the shortlisted cluster head terms (often 20–100 keywords, not 5,000).
3. Pick the cheapest source per data type:

Data	Cheapest source	When to pay more
Bulk volumes	DataForSEO Keywords Data (~$0.06 per 1,000 keywords)	Semrush or Ahrefs when you also need difficulty and SERP context in one place
Competitor keywords/traffic	Semrush (Semrush is strongest here)	—
Backlinks	Ahrefs Free for your own site	Ahrefs Lite/Standard for competitors; DataForSEO Backlinks for pay-per-use
Rank tracking	GSC (free, your own queries)	DataForSEO SERP Standard queue for competitor/locale tracking; Semrush or Ahrefs if you already pay
AI visibility	Manual frozen set + ai_visibility.py	Brand Radar or Otterly when you need weekly multi-engine tracking

The Main stack ranks by unattended automation first and by Georgian coverage, so for competitor keywords it puts SE Ranking and DataForSEO ahead of Semrush, and for AI visibility it puts ai_probe first.

1. Limit every call: row limits, single country/database, only needed fields, no historical data unless asked.
2. Estimate before running. Ahrefs free test targets show the unit cost in headers; Semrush has the free countapiunits balance check; DataForSEO lists prices per task. State the expected spend in the report. The credit guard defers any call above ask_above_pct to the end-of-run questions, so never stop mid-run to ask (references/autonomy.md).
3. Cache and snapshot. Store results in the project’s seo-work/ folder with the date; reuse within 30 days (volumes) or 7 days (SERPs) instead of re-pulling.
4. Use what’s included: plan MCP allowances (Semrush 50k units, Ahrefs plan units, included Brand Radar checks), Projects data and scheduled crawls before ad-hoc reports.
5. Seats: one paid seat per person who actually pulls data. Share exports, not logins (sharing logins usually breaks the terms).
6. Log the spend in SEO-WORKLOG.md (units/credits per task) to see which tool earns its cost.
7. Budget stacks:

Budget	Stack
$0/mo	GSC + Bing WMT + GA4 + PSI key + Ahrefs Free + Screaming Frog free + seo-suite. For unattended runs this is the default free stack
~$130–200/mo	Add one of: Ahrefs Lite ($129, links-heavy work) or Semrush Starter ($199, keyword and competitor-heavy work; the $139 SEO plan lists no MCP access, so a run can’t use it), plus a $50 DataForSEO wallet for bulk volumes and SERPs
~$500+/mo	Ahrefs Standard or Semrush Pro+, plus Screaming Frog licence, a DataForSEO wallet, and an AI tracker (Brand Radar or Otterly Standard) if AI visibility is a KPI

The default free stack costs $0 at every size. With paid tools set up, the Main stack needs no subscription for a small site (about $15–20 a month, pay per use). For a big site, Main is about $140–200 a month and Main + Cross-check, with Ahrefs Standard as the one subscription, about $400–710.