curl "https://api.hawkcrawl.com/v1/news?q=tesla&from=2024-01-01&to=2024-03-31&content=true" \
  -H "Authorization: Bearer YOUR_KEY"

And this comes back:

{
  "query": "tesla", "total": 50, "credits_used": 9, "cached": false,
  "articles": [{
    "title": "Tesla pauses German production after Red Sea attacks",
    "source": "The Guardian",
    "published_at": "2024-01-12T11:24:00Z",
    "resolved_url": "https://www.theguardian.com/technology/2024/jan/12/...",
    "content": "Tesla has said it will suspend most production at its...",
    "content_status": "ok"
  }]
}

What people build with it

Three shapes cover most of it, and the pricing is built so none of them is punished for the others:

• Retrieval for a model — pull the article body for a subject and hand it to your own pipeline. This is what most of the cost goes to, and it is the only part we meter heavily.
• Watching a subject — poll a query on a schedule and alert on what is new. Headlines are cheap to serve, so polling is cheap to do, however often you do it.
• Looking backwards — ask what was being written about something in 2014, or 2019, or last March. Date ranges are ordinary parameters, not a separate historical product.

We publish the denominator

Every news API in this market calls itself the best one and counts its sources. None of them publish a figure you can check. These are ours, measured on 18 September 2026 over 645 articles — thirteen unrelated topics, sampled in the last seven days and again in 2010, 2016 and 2021.

Full text is 77% overall, and the honest part is the spread: it runs from 40% to 100% depending on the subject. Housing, sport and technology are near the top, central-bank and climate coverage near the bottom, because that is where the paywalls are. The archive is not the weak half — June 2016 returned 86% against 72% for this week, since a ten-year-old article is rarely behind a wall.

What we cannot get is paywalls, publishers that refuse machine traffic, and pages that no longer exist. We name them rather than rounding them away, the response tells you per article which you got, and you are charged for none of them. The script that produces these numbers is in the repository, so the next run can disagree with this one.

If the API you were using has been retired

Microsoft shut down the Bing Search APIs on 11 August 2025 — the whole family, News included, free tier and paid. The official path forward is Grounding with Bing Search inside Azure AI Agents, which hands web context to an agent rather than returning structured results. If what you had was a call that returned news as JSON, that is not a replacement for it — the migration, field by field.

This is the same call, against a different index, with the archive and the article body attached. The documentation is the fastest way to tell whether it fits what you had.

What we do not return

No article summaries, no image URLs, no author names, no sentiment scores — the upstream index does not carry them and we will not invent fields we cannot fill. No web search, no image search, no other engines. One endpoint, done properly, is the whole plan for now.

If any of that is load-bearing for you, one of the other news APIs is a better fit, and it is better that you find that out here.

If you were going to use the RSS feed instead

Good instinct, and for one subject in one language it is the right call. We published the reference Google never wrote — four endpoints, nine query operators, sixteen language codes, an archive that reaches 1995, and the three walls that stop it being enough. Tested, not remembered.

Start with the documentation

Not the pricing page and not a demo. For a developer product the documentation is the product you are actually evaluating, so that is where we put the work: every parameter, the real response shape, what each call costs, and the failure numbers repeated wherever they matter.

Read the documentation Ask a question