> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentkey.app/llms.txt
> Use this file to discover all available pages before exploring further.

# News search for AI agents

> Headlines and articles by keyword from Google News via Serper, Brave's news index and Tavily's news topic — filtered by recency, region and language.

News is where an agent most often needs something the model has never seen:
what happened this morning, what a company announced an hour ago, how a story
is being covered in another country. Through AgentKey your agent searches
headlines with the same three calls it uses everywhere — `find_tools`,
`describe_tool`, `execute_tool` — choosing between Google News results through
Serper, Brave's own news index and Tavily's news topic, which can also return
the article text.

## What you can ask

* *"What are the top headlines on the Fed rate decision in the last 24 hours?"*
* *"Find news about lithium mining from UK sources this week."*
* *"Any news on Anthropic published in the past hour?"*
* *"Show me Japanese-language coverage of the Nikkei record high."*
* *"Search news published between 1 and 15 September about the EU AI Act."*
* *"Give me 50 recent news results about Tesla and group them by outlet."*
* *"Pull this week's news on the typhoon in Guangdong with the article text, and a short summary."*
* *"Get page two of the news results for this query."*

## What's covered

**Headlines by keyword**

* Google News results with title, link, source, date and snippet
* Brave news articles from its own index, with optional extra snippets and
  Goggles re-ranking
* Tavily's news topic, tuned for real-time updates, with each article's
  cleaned content as markdown or plain text on request and an optional
  generated answer

**Recency**

* Past hour, day, week, month or year on Google News
* The last 24 hours, 7 days, 31 days or 365 days, or a custom date range, on
  Brave
* A relative time range or explicit start and end dates on Tavily

**Region and language**

* Country and interface-language codes, plus a free-text location, on Google
  News
* Country, result language and interface language on Brave, with optional
  latitude / longitude, city or postal-code hints
* Domain include and exclude lists on Tavily to pin or drop specific outlets

**Volume and paging**

* A per-call result count on every engine; page numbers on Google News and an
  offset on Brave
* A POST variant of Brave news for long or multi-line queries

## How to use it

Pass the request in plain language and let the router pick the engine:

```text theme={null}
find_tools(q="latest news about Anthropic from the past week")
find_tools(q="Japanese coverage of the Nikkei record high today")
find_tools(prefix="search/serper")        # Google News results
find_tools(prefix="search/brave")         # Brave news index
find_tools(prefix="search/tavily")        # news topic with article text
describe_tool("<returned Provider/Operation>")
execute_tool("<Provider/Operation>", { ...params })
```

Always `describe_tool` before executing — the recency filter is `tbs` on
Google News, `freshness` on Brave and `time_range` or dates on Tavily, and
per-call costs differ.

> [!NOTE]
>
> **Building with code?** News search is part of the fixed-contract
> [`web.search`](/api-reference/scenario/methods/web-search) Scenario Method:
> set `type` to `news` and it returns a normalized result page with automatic
> engine failover. See [Scenario Methods](/api-reference/scenario/overview).

## FAQ

> **How fresh are the headlines?**
>
> Every call goes to the engine at request time. The recency filters narrow
> by the engine's own publication or discovery date, down to the past hour
> on Google News and the past 24 hours on Brave.

> **Which regions and languages?**
>
> Google News takes any Google country and interface-language code; Brave
> takes its country and language lists; Tavily is scoped by domain lists
> rather than by country. Queries can be written in any language.

> **Do I get the full article?**
>
> Google News and Brave return headline, source, date and snippet with the
> link. Tavily can include each article's cleaned content. For any other
> link, pass the URL to a [scrape tool](/capabilities/scrape/url-to-markdown)
> to read the page.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.