Best news API: 8 options for agents, ranked by price

A news API takes a keyword and returns matching articles as JSON, with a title, a source, a date and a link. If you build an agent, you usually want exactly that, one call per keyword, and no subscription for a job that runs a few hundred times. The hard part is picking a provider. Most news search endpoints read Google News, a few read their own index, and they differ in filters, fields and how a call is billed.

This post ranks the news search endpoints in the looot catalog by live price per call. You get what each returns, which source it reads, a decision guide, a prompt your agent can run and a FAQ.

Best news API: the short answer

Each bullet names a different need.

• AnyAPI, best for plain Google News headlines: one keyword, one request at a flat price.
• Serper, best for Google News results with a snippet and an image URL in one small response.
• DataForSEO, best for setting location and language per task, or getting the raw HTML of the results page.
• Cloro, best for picking the country and getting publisher, date and thumbnail for each article.
• Brave Search API, best for news from an index that is not Google News.
• NewsData.io API, best for the last 48 hours of articles plus crypto and market news endpoints.
• SerpApi, best when you already parse SerpApi JSON for other Google engines.
• Anysite, best for limiting results to one publisher, excluding words or requiring an exact phrase.

Prices below are read from the catalog when the page loads, one row per provider.

How we ranked these news APIs

We read the public looot catalog on 2026-11-04 and kept every provider with an endpoint tagged news_search that the blog can price. We ordered them by the price per call of that endpoint, cheapest first. DataForSEO has two such endpoints at the same price, so it gets one entry. Every endpoint here bills per call, not per article, so no per-result provider needed special placement. Cloro bills more for each extra page, so we ranked it on its first page.

No two prices tied. Had they tied, breadth of other endpoints in the catalog would have decided. We did not run these endpoints for this post. The catalog publishes no measured success rate or speed for them, so we rank on neither. The live ranking with logos is on the news API ranking page.

Google News API or its own index: what each reads

Six entries return Google News results: AnyAPI, Serper, DataForSEO, Cloro, SerpApi and Anysite. They differ in fields and filters, not in the source. Two read something else. Brave returns articles from Brave's independent index, so the article set is not the same as Google's. NewsData.io's description never mentions Google. It returns NewsData.io's own feed of articles from the past 48 hours. If you need a second opinion on coverage, pair one of each.

1. AnyAPI, best for plain headlines at a flat price

The endpoint anyapi-run-google-news searches Google News by keyword and returns fresh articles with headlines, sources, links and publish times as JSON. The catalog says the price is flat per request, the same whatever the number of results.

Best for

• A headline check across a list of brand names.
• Agents that only need title, source and link.

Strengths

• One flat price per request, so the bill is calls times price.

Limits

• The catalog lists no snippet field.
• The description says one lane serves this API, so a failed attempt has no fallback.

Does not fit a job that needs article summaries.

2. Serper, best for snippets and images in a small response

The endpoint serper-news returns Google News results with title, link, snippet, date, source and imageUrl for each article.

Best for

• A news digest that shows a short summary under each headline.
• Cards that need a thumbnail.

Strengths

• Snippet and image arrive in the same call as the headline.
• Serper also serves Google organic, Scholar and image results, so one provider covers several searches.

Limits

• The description lists no country, language or date filter. Read the endpoint inputs before you rely on one.

Does not fit a search that must stay inside one publisher.

3. DataForSEO, best for location, language and raw HTML

DataForSEO has two news_search endpoints. dataforseo-google-news-advanced returns Google News results live, with titles, sources, dates and snippets. dataforseo-google-news-html returns the raw HTML of the same results page. Both take a list of tasks, each with a keyword, a location_name and a language_code.

Best for

• Checking how coverage of one keyword differs by country.
• Keeping the raw page for your own parser.

Strengths

• Location and language are set per task.
• The HTML variant keeps what the parsed fields leave out.

Limits

• The request is a list of task objects, which takes more setup than a single keyword.
• The catalog says the charge follows DataForSEO's reported cost, so read the cost on each run.

Does not fit an agent that wants to send one string and get a list back.

4. Cloro, best for country, publisher and thumbnail

The endpoint cloro-monitor-google-news returns parsed Google News articles for a keyword from a chosen country. Each article has title, link, snippet, publisher, publication date and thumbnail.

Best for

• Tracking coverage of a launch in one country.
• Pages that show the publisher and a thumbnail next to each headline.

Strengths

• The catalog says a failed extraction is not charged.
• Results run up to 10 pages per call.

Limits

• Cloro bills 5 credits for page 1 and 2 more for each extra page, so a deep search costs more than the live row shows.

Does not fit a cheap, wide sweep over many keywords at several pages each.

5. Brave, best for a second news index

The endpoint brave-search-res-news-search returns news articles for a query from Brave's independent index. You send q, and country, search_lang, freshness and count refine it.

Best for

• Cross-checking Google News coverage with a different source.
• Recent-only searches through the freshness setting.

Strengths

• Country, language and freshness are explicit inputs.

Limits

• It is not Google News, so the article set will differ.

Does not fit someone who must match what Google News shows.

6. NewsData.io API, best for the last 48 hours and topic feeds

The endpoint newsdata-io-1-latest returns the most recent articles from the past 48 hours. It uses one NewsData.io credit per call. NewsData.io also has catalog endpoints for crypto news, financial-market news tagged with stock tickers, article counts and a list of news sources.

Best for

• A morning sweep of yesterday's coverage.
• Crypto or market news, through its other endpoints.

Strengths

• The sources endpoint returns source ids you can pass back to narrow a search to one domain.

Limits

• It takes no from_date or to_date, so it cannot search older coverage.
• The description says free-plan articles arrive with a 12-hour delay. It does not say which plan the looot key uses, so check the dates in the response.

Does not fit research into last month's coverage.

7. SerpApi, best when you already use its JSON

The endpoint serpapi-google-news returns raw SerpApi JSON with news_results, each with title, link, source, date and thumbnail. search_metadata.status reads Success when the search ran.

Best for

• Agents that already parse SerpApi responses for other engines.
• Combining news with Google Trends or Bing results from the same provider.

Strengths

• SerpApi lists Google, Bing, Yahoo, Yandex, Baidu, DuckDuckGo and Naver search in the catalog.

Limits

• The response is SerpApi's raw format, so your agent picks the fields out itself.

Does not fit a high-volume headline job where price per call decides.

8. Anysite, best for publisher and phrase filters

The endpoint anysite-api-google-news-articles-search searches Google News by keyword. exact_phrase, exclude_words, website and period narrow it, and count sets how many articles come back. Each article has title, source, source domain, publish date, snippet and link.

Best for

• Finding every mention of a product on one publisher's site.
• Excluding a common word that drowns the real results.

Strengths

• The most filters of any entry here.
• A second endpoint returns Google News topic sections such as BUSINESS or TECHNOLOGY.

Limits

• The website filter takes one publisher per call.

Does not fit a cheap, unfiltered sweep.

How to choose

What do you need back?

If you only need headline, source and link, start with AnyAPI. For a snippet and an image, use Serper or Cloro. For the raw page, use DataForSEO's HTML endpoint. For publisher and phrase filters, use Anysite. The Google News API use case lists each endpoint's filters side by side.

Google News or another source?

Six entries read Google News. For a different index, use Brave. For a 48-hour feed with crypto and market extras, use NewsData.io.

How many keywords do you run?

At a few dozen keywords, fields matter more than price. At thousands of calls a day, read the live table and the billing notes. Cloro charges per page, and DataForSEO follows its own reported cost. Web search API pricing shows the same arithmetic for web search.

Does it need to run on a timer?

A daily brand check needs scheduled runs, not available yet, so run it by hand each morning.

Try it

1. Connect your agent to looot once with the setup line below.
2. Ask it to search the catalog for news articles for a keyword. It returns each endpoint with its inputs and price.
3. Run two providers on the same keyword and compare the article lists.
4. Keep the one that returns the fields you need. Browse every endpoint in the public catalog.

To try one search for free first, the free Google News search tool returns one page of headlines as a table.

What this post does not cover

It covers keyword search, one call at a time. It does not cover full article text, sentiment or archives older than what each endpoint returns. It makes no claim about coverage or freshness, because the catalog publishes no measured rate. Providers in the live catalog that are not in the blog's provider snapshot, dated 2026-10-12, are left out until the snapshot is refreshed.

Last checked 2026-11-04 against the public catalog at https://api.looot.ai/v1/public-catalog. Author: the looot team.