How do I monitor the news without paying for the same story twice?

Q: How do I monitor the news without paying for the same story twice?

one wire story, billed once instead of forty times search news for 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 the set: ten real articles across CNBC, Bloomberg, Seeking Alpha, the Economic Times, a Japanese wire and an Argentine daily the key: every row carries a story_id, so the syndicated copies of one story can be grouped before you read them the spread: the same ten articles cost $0.0009 here and $0.0600 on the most expensive endpoint in the aisle two schema traps: the 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

Company News APIs

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

context.dev$0.0009

akta$0.0100

apollo$0.0500 · catalog

opoint$0.0600

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

akta7,341

opoint880

cloronot reported

surf0 on topic

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

context.dev · 5 releases3 were one release

opoint · 10 results5 were one wire

akta · 10 resultsnone observed

context.dev · 300 rows5 of 300

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

context.dev$0.09

apollo · catalog$0.50

akta$0.505

opoint · catalog$0.60

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

Company News

Try it with your agents

How-to guide

Set up Monid.

Pick what you need. Say it in one sentence.

Everything written about a company

Watch a topic, not a company

Sentiment and a summary on every headline

How much of a page is one story

A thousand articles, not ten

When the language field is wrong

Read more

Bing Search API retired: the alternatives

When the site will not let you in

Cheapest source first, then fall through