Search API · fused sources · one call

Cast wide.
Haul it in.
Sort the catch.

Your agent asks once. Trawlia queries every source that serves the topic in parallel, fetches the pages, strips the navigation and cookie banners, ranks what is left against the question, and hands back clean Markdown with citations.

Start free Read the docs

Every response says who answered

General topic · 12 Aug 2026

Four sources sit behind the API, and the topic decides which of them are asked: Bing and DuckDuckGo answer general, Google News and Yahoo News answer news. Three of the four are HTML scrapers, so every response carries a per-provider report. You can always tell a complete answer from a thin one, and an empty result array only ever means the web had nothing to say.

bing returned 10 304 ms

duckduckgo blocked 412 ms

A general search asks these two. One row is marked because DuckDuckGo soft-blocks plain requests and the browser tier that gets past it is not switched on, so today that search runs on one source and says so, rather than quietly returning a thinner answer than you asked for.

One call, not a pipeline

Without this, every agent rebuilds the same thing: run a search, fetch each result, parse the HTML, throw away the chrome, decide which passages matter, keep the token count sane, and handle the three sites that block you.

Trawlia does that and returns the part your model actually needs.

curl https://api.trawlia.co/v1/search \
  -H "authorization: Bearer twl_..." \
  -H "content-type: application/json" \
  -d '{
    "query": "what changed in the EU AI act",
    "search_depth": "advanced",
    "include_answer": true
  }'

What you get back

Clean Markdown

Readability pulls the article out of the page and it comes back as Markdown, headings and lists intact. Roughly a tenth of the tokens of raw HTML.

Ranked passages

Results from every engine that answered are fused by rank, so a page two providers agree on rises above any one provider's favourite. Then passages are scored against the question, not the page.

Cited answers

Ask for an answer and you get one, with citations validated against the sources actually supplied. A hallucinated reference is dropped before it reaches you.

On handling other people's pages

We identify ourselves and obey robots.txt

Requests go out as TrawliaBot with a contact URL, and we answer to robots.txt. When a source refuses that, the fallback is a real browser, not a plain HTTP client pretending to be one.

That distinction matters. Driving an actual browser is a different thing from lying about what you are, and any operator who wants to refuse us by name still can. It is also why Google News gives you headlines and never article bodies: their robots.txt says no, so we do not.

Point an agent at it

Free tier, no card. The request shape mirrors what your agent code probably already speaks, so trying it is a base URL change.

Start free See pricing

Cast wide.Haul it in.Sort the catch.