# SEO & search tools for AI agents: 42 tools, $0.001 to $0.50 per call

Whizbrew has 42 SEO & search tools. One call that returns data costs $0.001 to $0.50; a call that finds nothing or fails costs $0. They are part of one catalogue of 893 tools behind one MCP server and one API key ([all groups](https://whizbrew.com/tools?a=4b2c7c7f07c1), [pricing](https://whizbrew.com/pricing?a=4b2c7c7f07c1)).

## Example calls

`maps.search` ($0.002): Search Google Maps for businesses and places: name, rating, review count, address, phone, website, categories and the map's type id, business status (operational or closed for good), accessibility, booking and ordering links, opening hours, coordinates and place id, in the order the map ranks them. Put the area in the words ("dentist berlin") or give latitude and longitude. 20 places per call by default, up to 60 with "limit" (a lead list 40 or 60 deep in one call); "offset" reads on from next_offset.

```
curl -s -X POST https://whizbrew.com/v1/call/maps.search \
  -H "Authorization: Bearer $WHIZBREW_KEY" -H "content-type: application/json" \
  -d '{"query":"coffee shop","latitude":30.2672,"longitude":-97.7431,"limit":5}'
```

`maps.place` ($0.002): One Google Maps place by its place id: rating, review count and how many reviews gave each star, address, phone, website, categories and type id, business status, accessibility, booking links, opening hours per day, what the listing says the place offers, and the few reviews the place page shows (maps.reviews reads the reviews in full, as far as Google shows them to a visitor).

```
curl -s -X POST https://whizbrew.com/v1/call/maps.place \
  -H "Authorization: Bearer $WHIZBREW_KEY" -H "content-type: application/json" \
  -d '{"place_id":"ChIJRYqTNEZawokRuSMWeyyinyI"}'
```

`maps.rank_at` ($0.003): Where a business ranks on Google Maps for a keyword, searched from one coordinate: its rank (left out when it is not in the places read), the matching place and the five places on top. Reads the top 20; "depth" 40 or 60 reads further.

```
curl -s -X POST https://whizbrew.com/v1/call/maps.rank_at \
  -H "Authorization: Bearer $WHIZBREW_KEY" -H "content-type: application/json" \
  -d '{"query":"pizza","business":"joe\u0027s pizza","latitude":40.6943,"longitude":-73.9903}'
```

Over MCP the same call is `call_tool` with `{"tool": "<name>", "arguments": {…}}`.

## Every tool in this group

| Tool | What it returns | Price per call that returns data | Per 1,000 calls |
|---|---|---|---|
| `maps.search` | Search Google Maps for businesses and places: name, rating, review count, address, phone, website, categories and the map's type id, business status (operational or closed for good), accessibility, booking and ordering links, opening hours, coordinates and place id, in the order the map ranks them. Put the area in the words ("dentist berlin") or give latitude and longitude. 20 places per call by default, up to 60 with "limit" (a lead list 40 or 60 deep in one call); "offset" reads on from next_offset. | $0.002 | $2 |
| `maps.place` | One Google Maps place by its place id: rating, review count and how many reviews gave each star, address, phone, website, categories and type id, business status, accessibility, booking links, opening hours per day, what the listing says the place offers, and the few reviews the place page shows (maps.reviews reads the reviews in full, as far as Google shows them to a visitor). | $0.002 | $2 |
| `maps.rank_at` | Where a business ranks on Google Maps for a keyword, searched from one coordinate: its rank (left out when it is not in the places read), the matching place and the five places on top. Reads the top 20; "depth" 40 or 60 reads further. | $0.003 | $3 |
| `maps.rank_grid` | Local rank grid: where a business ranks on Google Maps for a keyword at every point of a 3x3 or 5x5 grid around a coordinate, with the average, best and worst rank, the share of points in the top 3 and the rivals seen most. One map read (top 20) per point, 25 points at most per call; the price is per call, whatever the grid size. | $0.02 | $20 |
| `maps.reviews` | Reviews of one Google Maps place by its place id: rating, text (and Google's translation when the review is in another language), language, date, the reviewer's display name, the owner's reply, photo links and rated aspects, with the place's rating, review count and how many reviews gave each star. Depth, honestly: Google shows a visitor who is not signed in only the 5 reviews on top of each order and no further pages. "sort" picks the order (newest, relevant, highest, lowest); "all" reads the four orders and returns up to 20 different reviews, newest first, each marked with the orders it was seen in. reviews_read and cut_by_source say how much of the place's list this is. Repeats are answered from a cache for 10 minutes. | $0.003 | $3 |
| `maps.reviews_summary` | The review picture of one Google Maps place in one cheap call: rating, review count, how many reviews gave each star, the themes Google itself names for the reviews (with how many reviews mention each) and the newest reviews (up to 5, with text, date, reviewer display name and owner reply). For more reviews of the place use maps.reviews with sort all. Repeats are answered from a cache for 10 minutes. | $0.002 | $2 |
| `trends.google_interest` | Google Trends search interest for a search term: the curve over time (0 to 100, 100 = the peak of the period) and the regions where it is searched most. Up to 5 comma-separated terms are compared on one scale. Values are relative, not search counts. A term with too few searches answers with empty lists. Answers are cached for 10 minutes. | $0.002 | $2 |
| `trends.google_related` | Google Trends related searches for one search term: the top queries people also search (0 to 100) and the rising ones with their growth ("+250%", or "Breakout" for a query that barely existed before). Up to 25 of each. A term with too few searches answers with empty lists. Answers are cached for 10 minutes. | $0.002 | $2 |
| `keywords.expand` | Long list of keyword ideas around a phrase: the search box suggestions for the phrase and for the phrase followed by each letter, without duplicates, most relevant first. depth fast makes 11 upstream reads (about 100 keywords, 1 to 2 seconds), depth full makes 27 (up to 300 keywords, about 4 seconds). relevance is the engine's own score for what was typed (google only), not a search volume. If some reads fail the answer is partial and calls.failed says how many. | $0.003 | $3 |
| `keywords.questions` | Questions and comparisons people type about a phrase: the search box suggestions for the phrase with a question word in front (how, what, why, when, where, which, who, can, is, are, does, best) and with "vs" and "for" behind, grouped by that word. Makes up to 14 upstream reads (1 to 2 seconds). Own word lists for the languages en, de, fr, es and pt; any other language is asked with the English words. A suggestion is listed under the word it was found with and may not contain it. If some reads fail the answer is partial and calls.failed says how many. | $0.003 | $3 |
| `seo.audit_url` | An on-page SEO audit of one URL, from our own fetch of the page (no script is run, so text and links that only appear after JavaScript are not seen). Checks: HTTP status and redirect chain (up to 5 hops), HTTPS, download time and HTML size, title and meta description (length in characters), heading counts and outline order, canonical (self, other, cross_domain, missing), noindex and nofollow from robots meta tags and the X-Robots-Tag header, whether robots.txt lets a general crawler fetch the URL, whether the first sitemap named in robots.txt lists it (one file of at most 1 MB; a sitemap index is not opened), hreflang entries (return links are not checked), lang, viewport, Open Graph and Twitter card tags, JSON-LD types and blocks that do not parse, links (internal means the same host), images without alt text, word count and http resources on an https page. Every problem is listed in plain words with the measured number and a severity; issue_counts is the count per severity (no score). Only the first 3 MB of HTML are read (truncated: true). Compression is not measured. An error page (404, 500) is still an answer. | $0.001 | $1 |
| `search.web` | Search the open web. By default several engines that answer without a key are read at the same time (Excite or Yahoo! JAPAN, Bing, NAVER, Daum, Seznam) and merged into one list: one entry per page, each with the engines that returned it and its rank at each, ordered by a combined score ("ranking" says how it is computed). This is our own merged list of thin logged-out results pages, not Google results and not any one engine's ranking. site (one domain), exclude (domains to leave out) and time_range are sent to the engines that keep to them and site and exclude are checked on every result; "site:" typed in the query works the same. Results that mention no word of the query are left out, and an answer with nothing left is not charged. Page on with cursor (next_cursor of the previous answer): about 30 to 60 distinct results a round, up to 5 rounds. engines_used, engines_failed and engines_skipped are in every answer, with checked_at (when it was read), result_count and, per engine that states one, its own estimate of all matching pages (total_estimate); one engine failing does not fail the call. With engine set to one engine the answer is that engine's own results page (bing: first page only, ignores site:; seznam: Czech market; yahoo_jp and excite: Japan, one index, so the merged list asks one of the two and the other only when the first refuses; naver and daum: Korea), paged with page. No language or country choice for these engines. GOOGLE: engine google (Google's first 10 results), google_30, google_50 or google_100 give Google's own organic results in Google's order for one country and language, read at the moment of the call through a licensed results reader: each result with its position and link, plus the people-also-ask questions and related searches when Google shows them. The Google engines take query, country and language only (no site, exclude, time_range, cursor, limit or search operators such as site:). Google returns fewer results than asked when it groups them. Google hands out redirect links and frees them slowly: about a third of results carry the final address on a first call; the rest carry the site name (site) and a redirect link that opens the page (url_is_redirect). The same call a few minutes later carries the final addresses. A Google answer takes about 8 to 11 seconds. Price by engine: google $0.004, google_30 $0.012, google_50 $0.02, google_100 $0.04; every other engine $0.001. | $0.001; with `engine` = google: $0.004, google_30: $0.012, google_50: $0.02, google_100: $0.04 | $1 |
| `search.site` | The pages of one domain that search engines know: what Excite or Yahoo! JAPAN (one index: one of the two answers, the other when the first refuses), NAVER, Daum and Seznam return for site:<domain>, merged into one list without duplicates and paged with cursor. Add query to find pages of the site about something (e.g. pricing). Each result lists the engines that returned it. This is what these engines show a logged-out reader (about 50 to 100 addresses each at most), not a crawl, not their whole index and not Google's: a large site is never listed in full. For the pages a site lists itself use domain.sitemap. | $0.001 | $1 |
| `search.news` | News results for a search text from DuckDuckGo's news page: title, link, publisher, publication time, a snippet and a picture. One page of at most 30 items, newest and most relevant mixed as the engine orders them, United States edition in English; time_range narrows to the last day, week or month. Many links lead to a news portal's copy of the article, with the publisher named in source. For up to 100 items per query, other countries and languages, use news.search. | $0.001 | $1 |
| `search.rank_check` | Where a domain (or any of its subdomains) stands on the FIRST results page of one engine, for 1 to 10 keywords. These are NOT Google positions: the engines are bing (United States, at most 10 results, can return unrelated results for rare words), seznam (Czech), yahoo_jp (Japan) and naver (Korea). position is null when the domain is not among the results read; "checked" says how many results were read for that keyword. Up to 20 seconds for 10 keywords. GOOGLE: engine google gives the domain's Google position (top 100) for each keyword in one country, from a licensed SEO index, with the ranking page, the keyword's monthly searches and checked_at: when the index last read Google for that keyword (days to weeks ago; a modelled snapshot, not read now). position null = not in Google's top 100 then, or a keyword the index does not follow. Price: $0.032 with engine google, $0.003 otherwise. | $0.003; with `engine` = google: $0.032 | $3 |
| `keywords.difficulty` | An ESTIMATE (0 to 100) of how hard a keyword is to rank for, computed here, not taken from an SEO tool and not from Google's top 10: it reads the first results page of one engine (bing by default, United States, at most 10 results) and looks up how popular each result's domain is among the one million most visited sites. Many strong domains at the top give a high score. The answer returns the formula and every input (position, domain, rank). When the page the engine returned is not about the keyword there is no difficulty: warning says so, the rows are in unrelated_results and a single-keyword call is not charged. Give "keywords" to score up to 10 keywords in one call (one results page each, about a second a keyword): the answer is then {keywords: [...]} with one row per keyword. scores both adds index_difficulty beside ours: the 0 to 100 keyword difficulty of a licensed SEO index for Google in one country (a modelled estimate too, from other inputs; null for a keyword the index does not have). Price: $0.03 with scores both, $0.003 otherwise; when the index is not available a scores both call fails and is free. | $0.003; with `scores` = both: $0.03 | $3 |
| `search.images` | Image results for a search text: title, the image file, the page it is on and a thumbnail. Engine bing: 35 a page; engine duckduckgo: up to 100 a page, with each picture's width and height. Pages 1 to 5, United States, English. No size, colour or licence filter. A word the engine does not know gives loosely matching pictures instead of an empty list (then "warning" is set). | $0.001 | $1 |
| `search.videos` | Video results for a search text, from one of three engines. bing (default): mostly YouTube and TikTok, with length, rounded view count, age as printed ("3 weeks ago", not a date), site and channel; one page of at most 50, United States, English. duckduckgo: up to 60 a page, pages 1 to 5, with exact view count, publication date, channel and description. yandex: about 17 videos, first page only, strong for Russian-language and other video sites next to YouTube; title, link, site, description and thumbnail, no length, views or date. | $0.001 | $1 |
| `search.shopping` | Product results for a search text: title, price in US dollars as printed (one seller's offer, not the lowest price on the web), seller, rating and picture. One page of at most 36, no paging, United States only. The link is the engine's product page, not the merchant's. An unknown word gives unrelated products (then "warning" is set). | $0.0015 | $1.50 |
| `backlinks.summary` | Domain authority and backlink profile of a domain: authority (0 to 100, our own score from the domain's rank in the crawl on a log scale: 100 = the most central domain; not Moz DA or Ahrefs DR), how many domains link to it (referring domains), how many it links to, and its harmonic centrality and PageRank with its rank for each. No traffic estimate. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. totals index adds index_totals beside these numbers, from a licensed page-level link index with its own crawler, read now: all backlinks, referring pages, referring domains, nofollow referring domains, broken backlinks, the split by ending, kind of link and country, each labelled with its source and date. The two sources count different things and will not match. Price: $0.05 with totals index, $0.001 otherwise; when the index is not available a totals index call fails and is free. | $0.001; with `totals` = index: $0.05 | $1 |
| `backlinks.referring_domains` | The domains that link to a domain, best ranked first, page by page with offset. sort: rank (most central first), rank_low (least central first) or pagerank. For a domain with over 2,000 referring domains the default order leaves out the referrers among the web's 1,000 most central sites (the same giants link to every large site); skip_top changes that number, 0 lists everything. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.002 | $2 |
| `domain.compare` | Compare up to 10 domains side by side in one call: our domain authority score (0 to 100), the exact number of referring domains and the rank in the link graph, and each domain's popularity rank among the one million most visited sites. Strongest authority first. The numbers of backlinks.bulk_summary and domain.rank joined, so nothing has to be stitched. Authority is our own score, not Moz DA or Ahrefs DR, and there is no traffic estimate. | $0.005 | $5 |
| `backlinks.outgoing` | The domains a domain links to, best ranked first, page by page with offset. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.002 | $2 |
| `backlinks.gap` | Link gap: the domains that link to one or more given domains (your competitors) and not to another (you), best ranked first. match all keeps only those that link to every given domain. Without exclude it is the shared or combined referrers. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.003 | $3 |
| `backlinks.top_domains` | The best-ranked domains of the crawl: of the whole web, of one ending (tld), or those whose name matches a pattern (searched among the one million best-ranked). Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.001 | $1 |
| `backlinks.bulk_summary` | Domain authority and backlink profile of up to 100 domains in one call, to compare sites: authority (0 to 100, our own score from the rank in the crawl; not Moz DA or Ahrefs DR), referring domains, linked domains and both authority ranks for each. No traffic estimate. Free when the crawl knows none of them. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.005 | $5 |
| `backlinks.links_between` | Does a domain link to other domains, and do they link to it: for one domain and up to 100 others, links_to (the domain links to the other) and linked_from (the other links to the domain). Use it to check which prospects, partners or directories already link to a site, or whether two sites link to each other. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.002 | $2 |
| `backlinks.shared_outgoing` | The domains that one or more given domains link to, best ranked first: with match all, the sites every one of them links to (the directories, partners and sources a group of competitors has in common); with exclude, only those that another domain (you) does not link to. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.003 | $3 |
| `backlinks.overlap` | How the referring domains of 2 to 5 domains overlap, as counts: how many link to each one, to only that one, to every one of them, to each pair, and to exactly 1, 2, 3 of them. The numbers behind a link-overlap chart; backlinks.gap lists the domains themselves. Very large domains are counted over their best-ranked referrers only (complete is then false). Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.003 | $3 |
| `backlinks.referring_tlds` | Which domain endings (com, de, co.uk counts as uk) a domain's referring domains have, most frequent first, with the share of each and its best-ranked example: the country and kind profile of a link profile. Counted over the 5,000 best-ranked referring domains at most. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.002 | $2 |
| `backlinks.tlds` | The domain endings of the crawl with how many domains each has, most first: how large each country's or ending's web is in the link graph. Use an ending with backlinks.top_domains to list its best-ranked domains. Domain level, from an open web crawl of August to October 2026: one link per pair of domains, no pages or anchor texts. Each answer names its crawl in source. A domain the crawl does not know is free. | $0.001 | $1 |
| `seo.ranked_keywords` | The keywords a domain ranks for in Google's top 100 in one country: for each keyword the position, the ranking page, monthly searches, cost per click, competition, difficulty (0 to 100), search intent, the estimated monthly visits it brings and when the index last read it. Sorted by estimated visits (or position, or search volume); page on with offset. summary has the domain's totals: ranking keywords, estimated organic traffic and how many keywords stand in each position range. A domain the index does not know is free. From a licensed SEO index: the numbers are modelled estimates as the index last read each keyword (as_of), not measurements, and other SEO tools give other numbers. Price by rows: 10 $0.027, 25 $0.03, 50 $0.036, 100 $0.048. | $0.03; with `rows` = 10: $0.027, 25: $0.03, 50: $0.036, 100: $0.048 | $30 |
| `seo.domain_overview` | Estimated organic search traffic of 1 to 10 domains in one country: for each domain the estimated monthly visits from Google's organic results, how many keywords it ranks for, and the same for paid search. One domain alone also gets its keyword counts by position range (top 1, 2 to 3, 4 to 10, 11 to 20, 21 to 50, 51 to 100), the value of its traffic in ad money and how many keywords are new, up, down or lost. Domains the index does not know are named in not_found; a call that finds none is free. For authority and referring domains use backlinks.bulk_summary. From a licensed SEO index: the numbers are modelled estimates as the index last read each keyword (as_of), not measurements, and other SEO tools give other numbers. $0.027 a call. | $0.027 | $27 |
| `seo.competitors` | Organic search competitors of a domain: the domains that rank in Google's top 100 for the same keywords in one country, most shared keywords first, each with the number of shared keywords, its average position on them, its own keyword count and estimated organic traffic, and the visits the shared keywords bring it. The web's largest sites (Wikipedia, YouTube and the like) are left out unless include_giants is true. A domain the index does not know is free. From a licensed SEO index: the numbers are modelled estimates as the index last read each keyword (as_of), not measurements, and other SEO tools give other numbers. Price by rows: 10 $0.027, 25 $0.03, 50 $0.036. | $0.027; with `rows` = 10: $0.027, 25: $0.03, 50: $0.036 | $27 |
| `keywords.volume` | Search volume for up to 20 keywords in one country, in one call: monthly Google searches (the average of the last 12 months), the 12 months one by one, the change against last month, quarter and year, cost per click and the advertisers' low and high bids in US dollars, ad competition (0 to 1 and low, medium or high), difficulty (0 to 100) and search intent. Keywords the index does not have are named in not_found; a call that finds none is free. From a licensed SEO index: the numbers are modelled estimates as the index last read each keyword (as_of), not measurements, and other SEO tools give other numbers. $0.029 a call. | $0.029 | $29 |
| `keywords.ideas` | Keyword ideas with their numbers: keywords of the same topics as 1 to 5 seed keywords, most relevant first (or most searched first), each with monthly Google searches, cost per click, competition, difficulty (0 to 100) and search intent for one country. match phrase keeps only keywords that contain the seed words. total says how many the index has. For search-box suggestions without numbers (cheaper) use keywords.expand. From a licensed SEO index: the numbers are modelled estimates as the index last read each keyword (as_of), not measurements, and other SEO tools give other numbers. Price by rows: 10 $0.027, 25 $0.03, 50 $0.036, 100 $0.048. | $0.03; with `rows` = 10: $0.027, 25: $0.03, 50: $0.036, 100: $0.048 | $30 |
| `backlinks.pages` | Page-level backlinks of a domain or of one page: the pages that link to it, strongest first, each with the linking page's address and title, the page it links to, the anchor text, whether the link is nofollow, the kind of link (anchor, image, redirect), when the link was first and last seen, and 0 to 100 link scores of the linking page and domain (the index's own, not Moz DA or Ahrefs DR). By default one link per linking domain; per_domain all lists every link. total is the number of linking domains (or of links). From a licensed page-level link index with its own crawler, live links only: our own backlinks.* tools are domain level from an open crawl and count differently. A target the index does not know is free. Price by rows: 10 $0.049, 25 $0.05, 50 $0.052, 100 $0.056. | $0.05; with `rows` = 10: $0.049, 25: $0.05, 50: $0.052, 100: $0.056 | $50 |
| `ai.answer` | What an AI assistant answers to a question, asked now with web search on: the answer text, the links it cites, and for each brand name or domain you give whether it is mentioned, how often, where first and which cited links are its own. model picks the family: perplexity, chatgpt or gemini. This is the family's API answer; its consumer app can answer differently. Price by model: perplexity $0.015, chatgpt $0.04, gemini $0.1. | $0.015; with `model` = perplexity: $0.015, chatgpt: $0.04, gemini: $0.10 | $15 |
| `ai.brand_check` | Brand visibility in AI answers: asks one model family up to 5 questions at once (web search on) and reports for each brand name or domain in how many answers it is named and cited, with every answer, its links and the mention details. One price for up to 5 questions. Price by model: perplexity $0.075, chatgpt $0.2, gemini $0.5. This is the family's API answer; its consumer app can answer differently. | $0.075; with `model` = perplexity: $0.075, chatgpt: $0.20, gemini: $0.50 | $75 |
| `content.brief` | A content brief for a keyword in one call: the pages that rank for it (our merged web search, not Google's own page), and for the first of them the word count of the main content, the publication date and the h1 to h3 outline; the median length with a suggested range; the words and phrases the headings of several of these pages share (what a page on the topic is expected to cover); and the questions people type about the keyword. Pages are read as the server sends them, once each. Takes 5 to 15 seconds. Free when the search finds nothing. | $0.01 | $10 |
| `content.gap` | What a competitor's site has pages for and yours has not, from the two sitemaps: the sections (first part of the path) the competitor fills and you lack, the sections where it has at least three times your pages, and the topics (words in its page addresses) on three or more of its pages and none of yours, each with an example page. Up to 1,000 page addresses per site are compared; addresses, not page contents. Free when the competitor's sitemap shows nothing you lack. | $0.004 | $4 |
| `seo.schema_markup` | Writes the structured data (JSON-LD, schema.org) search engines read for rich results, from plain fields: FAQ, article, product (price, availability, rating), organization, local business, breadcrumb, event, software app or video. Returns the JSON-LD, the script tag to paste, the recommended fields still missing and warnings (a rating without a review count, a date that is not ISO). It writes markup for facts you give; it does not read your page (web.structured_data reads what a page has). | $0.001 | $1 |

Inputs, output shape, health and measured success rate of each tool: [https://whizbrew.com/v1/tools](https://whizbrew.com/v1/tools) and [https://whizbrew.com/status#seo](https://whizbrew.com/status#seo).

## Start

- **An agent on its own**: read [https://whizbrew.com/agents.md?a=4b2c7c7f07c1](https://whizbrew.com/agents.md?src=body-text&a=4b2c7c7f07c1). It asks for a free quote (`GET https://whizbrew.com/v1/agent/guide?goal=<the task in plain words>`), then one request (`POST https://whizbrew.com/v1/agent/start?a=4b2c7c7f07c1`) gives it a key that works at once with the free credit, with no email and no sign-up form. Its human claims the account later.
- **A chat app** (ChatGPT, Claude, Gemini): add the MCP server `https://whizbrew.com/mcp` as a custom connector and sign in. No key to copy.
- **An agent with a shell** (Claude Code, Codex, Cursor): read [https://whizbrew.com/llms.txt?a=4b2c7c7f07c1](https://whizbrew.com/llms.txt?src=body-text&a=4b2c7c7f07c1). The agent starts a pairing with one request, its user approves with one click, and the agent holds its own key.
- **Plain HTTP**: create an account at [https://whizbrew.com/app#signup](https://whizbrew.com/app#signup), make a key, then `POST https://whizbrew.com/v1/call/<tool>` with `Authorization: Bearer <key>`.
