How-to · 3 steps
How to monitor the news
with Monid.
Give this to your agent
$set up https://monid.ai/SKILL.md, then watch the news for <company> and show the cost of each step
and let it take it from there.
#JobWhat came backCost
1Resolve the name25 candidates, ours on top$0.00 ✓
2Ten articles30 days, a story_id on each$0.0009 ✓
3The same tensentiment, summary, entities$0.01 ✓
4A topic instead7,341 matches, no company$0.01 ✓
5The expensive one880 matches, 5 of 10 a wire$0.06 ✓
6The narrative onepromised 20, sent 2$0.055 ✓
7The off-scope onecrypto stories, policy query$0.006 ✓
8The localised oneonly the dates went Chinese$0.004 ✓
9A real missno articles, nothing billed$0.00 ✓
10Totalthirteen runs, eleven billed$0.1768 ✓
one company · one topic · an 11x spread on the same queryagent running$0.1768
Real run, 2026-09-25. A public company, public coverage.
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.
Everything written about a company
Say this to your agent
>search news for <company> in the last 30 days with context.dev on monid, give me ten articles with the date, source and url, and keep the story_id on each one
What the agent runsdone
context.dev /news/search$0.00009 / result
akta /v1/news$0.0005 / result
apollo /news_articles/search · not run$0.05
opoint /search$0.06
surf /project/ai-news · not run$0.012
What came backone company, one topic · 2026-09-25
Ten articles, same company, same 30 dayswhat the call billed
the setten real articles across CNBC, Bloomberg, Seeking Alpha, the Economic Times, a Japanese wire and an Argentine daily
the keyevery row carries a story_id, so the syndicated copies of one story can be grouped before you read them
the spreadthe same ten articles cost $0.0009 here and $0.0600 on the most expensive endpoint in the aisle
two schema trapsthe price list calls the field headline and the wire sends title; and the rival's schema says a bare company name is not accepted, which is false, it returns the same ten as the uuid
$0.0009ten articles
Watch a topic, not a company
Say this to your agent
>search news for '<topic>' in the last 30 days with akta on monid, no company binding, and tell me how many matched in total before you show me the ten
What the agent runsdone
akta /v1/news$0.0005 / result
opoint /search$0.06
cloro /google/news$0.008
apify /data_xplorer/google-news-scraper-fast · not run$0.006 / result
surf /search/news$0.006
What came backone company, one topic · 2026-09-25
Same topic, same window, what each one sawtotal matches reported
the number7,341 articles matched the topic in 30 days, against 880 on the endpoint that cost six times more
what that isa difference in index size and matching, not a difference in the world; neither total is the truth
the wrong shelfone endpoint answered a policy query with ten crypto stories, billed $0.0060, and never signalled that it was off scope
the habitask for the total first, then the page, so a suspiciously small total gets caught before you build on it
$0.01ten of 7,341
Sentiment and a summary on every headline
Say this to your agent
>get news for <company> with akta on monid and keep the sentiment label, the numeric score and the ai summary on each article. do not swap in a cheaper news endpoint to save money and do not call a second one for sentiment: context.dev returns the same articles for a tenth of the price, carries no sentiment field at all, and still answers 200
What the agent runsdone
akta /v1/news$0.0005 / result
finviz /get_stock_news_sentiment$0.01
api.itsgloria.ai /news-ticker-summary · not run$0.0341
marketbeat /get_stock_headlines$0.01
What came backone company, one topic · 2026-09-25
what arrivesa sentiment label, a numeric score, an AI summary, the publisher, countries and entity tags, all in the one response
the swap that 200sthe endpoint that is eleven times cheaper returns the same articles with no sentiment field anywhere, and nothing in the response says so
read the score rightsentiment_score is polarity, not confidence: Negative came back as 0.0038 and Positive as 1, so reporting it as certainty inverts the meaning
the case trapthe response says Positive and the filter enum wants positive, so a value read out of one response will not go back into the next query
the ticker pairtwo market endpoints billed $0.01 each for about ten and seven headlines; one tags sentiment and the other does not, and both are snapshots rather than archives
$0.01ten, scored and summarised
How much of a page is one story
Say this to your agent
>before you show me the news for <company>, group the results by story_id and tell me how many distinct stories there were against how many rows i paid for. do not expect this to save money: nothing lets you ask for distinct stories, so the rows bill either way
What the agent runsdone
context.dev /news/search$0.00009 / result
opoint /search$0.06
akta /v1/news$0.0005 / result
opoint /search-by-ids · not run$0.06
What came backone company, one topic · 2026-09-25
How much of a page is one storyrows that collapse, billed in full
the small sampleone wire story came back five times in ten opoint results, and one press release three times in five, each copy billed as its own row
the big sampleacross 300 rows of general coverage the collapse was 295 distinct stories, 1.7%, so the press-release page is the worst case and not the rule
why measure it anywaycontext.dev exposes a story_id to group on and opoint exposes nothing, so on one of them you cannot even tell how much you paid twice for
no discount either waythere is no way to push the grouping down to the endpoint, so this is a cleanup step, not a saving; akta's group_articles was billed and changed nothing
$0.00client side, and it saves nothing
A thousand articles, not ten
Say this to your agent
>pull 1,000 articles about <company> with context.dev on monid and page through it with the cursor. after each page tell me the run id and the rows so far: the file it writes carries no cost and this endpoint reports no total, so work the price out from the rows yourself
What the agent runsdone
context.dev /news/search$0.00009 / result
akta /v1/news$0.0005 / result
opoint /search · not run$0.06 / 100
apollo /news_articles/search · not run$0.05 / 100
What came backone company, one topic · 2026-09-25
What 1,000 articles costsUS dollars
the cheap route$0.09 for a thousand articles; three pages of a hundred came back with zero overlap, monotonic dates and distinct cursors, so the paging is clean
what the extra buysakta is 5.6x more and bundles sentiment, a summary and entity tags; that is a real trade, not a worse price
the flat-fee trapsthe per-call endpoints charge the same for a page of two as for a page of a hundred, so small precise queries are punished there and free here
no denominatorthis endpoint reports no total match count, so "stop when it stops being worth it" has nothing to measure against; decide on the rows you have read
$0.09per 1,000 articles
When the language field is wrong
Say this to your agent
>search news about <topic> in <country> with akta on monid, then read the headline text yourself and tell me which articles disagree with their own language field. do not filter on that field: it reported english for a chinese-titled article and chinese for a korean one
What the agent runsdone
cloro /google/news$0.008
akta /v1/news$0.0005 / result
apify /data_xplorer/google-news-scraper-fast · not run$0.006 / result
tikhub /api/v1/toutiao/web/get_article_info · not run$0.0015
tinyfish /search · not run$0.00
What came backone company, one topic · 2026-09-25
the locale that did nothingthe same policy topic with the country set to CN and the interface language to zh-CN returned ten English-language global outlets with the dates rendered as 2 天前
the field that liesone endpoint reported EN for a Chinese-titled article and EN for a Czech one; another reported zh for a Korean headline and English for three Xinhua pieces
the country enumthe cheapest endpoint rejects cn outright: its list has ch for Switzerland and no code for China at all, and its search is bound to a named entity so a topic query returns zero rows for $0.00
what does workread the headline; and know that the countries field is the country of the event, not the outlet, which is why a US Senate story matched CN
$0.004ten results, one field you cannot trust
Step 3
Take the cheapest route with Monid.
Resolving the company is free. The articles are where the money is, and the first rung is a hundredth of the last one.
Resolve the name · $0.0000
Company search$0.00find the company
Resolve filters$0.00country and language ids
Today's articles · cheapest first
Context.dev news$0.0009
Cloro Google News$0.004
Surf news search$0.006
Akta news$0.01
Opoint search$0.06
Group by story_id
Collapse the copies$0.00
Score and summarise$0.01only if you need it scored
Back to you
rowread fromcost
1Context.dev news$0.0009
2Context.dev news$0.0009
3no coverage$0.00
4Akta news$0.02
5Context.dev news$0.0009
5 of 5 rows$0.0227
Say this to your agent
>watch the news for these companies every morning, cheapest source first, and show the cost per company: <paste names>
“cheapest source first”ten articles cost $0.0009 on the first rung and $0.0600 on the last one
“group by story_id”one wire story filled half a page of results and every copy was billed
“ask for the total first”one endpoint reported 7,341 matches and another 880 for the same topic
“read the headline yourself”the language field reported english for a chinese article and chinese for a korean one
One key. 1,700+ tools.
News is two aisles: 23 search endpoints and 4 tied to a company record.
Cheapest per articlecontext.dev · $0.00009 / result
News with sentimentakta · $0.0005 / result + $0.005
Google News, 50+ marketsapify · $0.006 / result
Google News by marketcloro · $0.008 / unit
Crypto feed and detailsurf · $0.006 to $0.012
Narrative signals and recapsgloria · $0.033 to $0.11
News on a company recordapollo · $0.05 / call
Global press, five ways inopoint · $0.06 / call
Toutiao articles and commentstikhub · $0.0015 / call
Web search, no chargetinyfish · $0.00 / call
+ 1,700 more across 55 providers
Browse the catalog →Read more
View all →
Bing Search API retired: the alternatives
When the story is not in a news index yet, searching the live web is the next call.
When the site will not let you in
A browser that signs in and clicks, for the outlet behind a paywall.
Cheapest source first, then fall through
The same route, applied to finding a person rather than a headline.