How-to · 3 steps
How to research an Amazon product
with Monid.
Give this to your agent
$set up https://monid.ai/SKILL.md, then tell me what actually goes wrong with <asin> before i buy it
and let it take it from there.
#JobWhat came backCost
1The one-stars20 reviews, one root cause$0.027 ✓
2The spread73% five-star, 8% one-star$0.00 ✓
3The rank#1 in its category, #22 overall$0.05 ✓
4The complaint65% say the motor dies$0.00 ✓
5The market48 rivals with price and rating$0.0072 ✓
6The empty onepromised rank, returned nothing$0.00 ✓
7All 2,014 reviewswhat it would have cost$2.72 ✓
8Totalnew spend, reusing the rest$0.0572 ✓
one product · 20 reviews that mattered · 48 competitorsagent running$0.0572
Real run, 2026-09-24. A public listing, named by category.
Step 1
Set up Monid.
One line. It installs the CLI and asks for an API key from app.monid.ai. New accounts start with $1.00.
Say this to your agent
>set up https://monid.ai/SKILL.md
Step 2
Pick what you need. Say it in one sentence.
Click a job. Paste the sentence, fill in the brackets.
What goes wrong with this product
Say this to your agent
>pull the one-star reviews for <asin> with apify on monid, filtered to one star, and leave the page count alone unless you need more: it defaults to one page of ten and asking for two doubles the bill. this endpoint bills per review whatever its schema says
What the agent runsdone
apify /axesso_data/amazon-reviews-scraper$0.00135 / review
apify /web_wanderer/amazon-reviews-extractor · not run$0.00135 / review
apify /delicious_zebu/amazon-product-details-scraper · not run$0.00375 / result
mrscraper /amazon/product$0.05
What came backone product · 2026-09-24
What a review pull costsUS dollars
two pages is not a capthe default is one page of ten reviews at $0.0135; asking for two pages billed $0.027, exactly double, so the page-count clause spends money rather than saving it
what it would have cost$2.72 to pull all 2,014 reviews on this one product
the filter is the pointone-star, filtered, is not a sample: it is the whole complaint set, and one page of it is the default
the schema is wrong about the priceinspect says one result is returned per input query and that there is no result limit, and both are false: one query returned twenty results and twenty billed units. Trust the run's cost over the schema here
free extraevery review carries the product-wide rating spread, so that job needs no second call
$0.02720 reviews at $0.00135
How the ratings actually split
Say this to your agent
>from those reviews, give me the rating distribution and tell me what share of buyers are unhappy. it is already in the response you paid for
What the agent runsdone
apify /axesso_data/amazon-reviews-scraper$0.00135 / review
mrscraper /amazon/product$0.05
What came backone product · 2026-09-24
Where the 239,000 ratings sitshare of ratings
the headline4.4 stars across 239,000 ratings, which reads as a safe buy
the tail8% one-star is roughly 19,000 people, which is a large absolute number
why it mattersthe average hides a specific failure that the one-star reviews name precisely
what the field is calledit is a review summary rather than anything named distribution, it gives percentages rather than counts, and it is repeated on every row, so an agent mapping rows to reviews drops it without noticing
the drifttwo sources nine days apart disagreed by 81 ratings, so treat the count as a snapshot
$0.00already in the reviews
Is it actually selling
Say this to your agent
>get the product detail for the amazon URL of <asin> with mrscraper on monid: price, rating and badges. it takes a url rather than an asin, it is a flat $0.05 however big the response, and it answers with a download link you have to fetch separately within the hour. do not count on a bestseller rank: on a second product there was none anywhere in the payload
What the agent runsdone
mrscraper /amazon/product$0.05
apify /delicious_zebu/amazon-product-details-scraper$0.00375 / result
apify /axesso_data/amazon-search-scraper$0.00015 / result
What came backone product · 2026-09-24
the ranknumber one in its own category and number 22 in the whole department
the price$8.88 against a $16.99 list price, with a choice badge and 20,000 bought last month
the failurethe cheaper actor that promises bestseller ranks returned an empty array twice, and billed nothing
the rank is not always thereon a second product there was no rank key, no bestsellers row in the technical details, and the phrase did not appear in two million characters of raw html; the best-seller field there is a badge boolean, not a rank
and it is a link, not a payloadthe response is a download link with about an hour to live, so saving the run to a file saves a few hundred bytes of pointer rather than the 2 MB of data
the sizethe flat $0.05 bought a 3.3 MB response, most of it raw html you will not read
$0.05flat, per product
Rank the complaints
Say this to your agent
>from the one-star reviews you already pulled, rank the complaint themes by how many reviews mention each one, and tell me what order they were sorted in. the sort defaults to most recent, so this ranks recent complaints rather than common ones unless you set it. do not call anything again
What the agent runsdone
apify /axesso_data/amazon-reviews-scraper$0.00135 / review
typesafe /judge · not run$0.042
What came backone product · 2026-09-24
What the one-stars actually sayshare of 20 reviews
the root cause13 of 20 one-star reviews describe the motor failing within one to six months of light use
the restparts detaching, weak from new, damaged on arrival, and a warranty that still charges return postage
what it costnothing: this is counting themes in text you already bought
the reada 4.4-star product with one specific, predictable failure is a different purchase than a 4.4-star product without one
$0.00computed, not called
What it is competing against
Say this to your agent
>search <keyword> on amazon with apify on monid and give me the rows with price, rating and review count, flagging any row whose price came back as zero. do not loop a detail endpoint over the results: the dearest is 333x the per-row price and even the cheapest is 25x
What the agent runsdone
apify /axesso_data/amazon-search-scraper$0.00015 / result
apify /delicious_zebu/amazon-product-details-scraper · not run$0.00375 / result
mrscraper /amazon/product · not run$0.05
litescrape /google/shopping · not run$0.00015
What came backone product · 2026-09-24
48 competitors, three waysUS dollars
the table48 competing products with price, list price, rating, review count and the sponsored flag
the price per rowabout $0.00015, which makes a whole category cheap to survey
not every row has a priceon a second keyword two of sixteen rows came back with a price of zero and were billed like the rest, so the promise of a price per row is about seven eighths true
the wrong waythe same 48 rows built one product at a time is $0.18 at best and $2.40 at worst
the gotchathe category filter was accepted, ignored, and billed: the results came back unscoped
$0.007248 rows at $0.00015
Step 3
Take the cheapest route with Monid.
Filter before you buy. Reviews bill per review, and a search row costs a fraction of a product page.
Search · $0.00015 a row
Search rows$0.007248 with price and rating
Google Shopping$0.00015the price elsewhere
Go deeper · cheapest data first, then judge it
Product details$0.0037
Reviews, filtered$0.027
Full product page$0.05
Judge the batch$0.042
Name the risk
Count the themes$0.00
Mail the digest$0.001one email · catalog rate
Back to you
rowjudged bycost
1Reviews, filtered$0.0272
2too few ratings$0.0001
3Full product page$0.0502
4Reviews, filtered$0.0272
5too few ratings$0.0001
5 of 5 rows$0.1047
Say this to your agent
>tell me which of these products has a failure pattern in its one-star reviews, cheapest source first, and show the cost per product: <paste asins>
“leave the page count alone”it defaults to one page of ten, so asking for two doubles the bill rather than capping it
“rows, not pages”48 competitors as search rows is $0.0072; as product pages it is $2.40
“read the whole response”the rating spread arrives inside the reviews, so it needs no second call
“check the filter took”a category filter was accepted, ignored and billed in full
One key. 1,700+ tools.
Ecommerce is one aisle.
Reviews, search, product actorsapify · from $0.00015 / result
One product, everything on itmrscraper · $0.05 / call
TikTok Shop productstikhub · $0.0015 / call
Google Shoppinglitescrape · $0.00015 / query
Read a product pagecontext.dev · from $0.0009
Judge a batch of reviewstypesafe · $0.042 / call
Find the competitorsexa · $0.002 to $0.01
Mail the digestagentmail · $0.001 / email
Narrate the findingselevenlabs · to $0.22 / call
Cut a clip from itminimax · $0.038 / s at 480P
+ 1,700 more across 55 providers
Browse the catalog →Read more
View all →
Claude for Amazon reviews
The same job as a loop: what the complaints say across a whole category.
Extract page content for RAG
When a listing is only on a page, reading the page is the next call.
The best social media scraping API in 2026
What buyers say off Amazon, and what each platform costs to read.