Skip to main content
Search finds pages from a query and returns them ranked, grouped by source. Pick the sources in sources; with no sources you get web results.

Make a request

Get an API key in the Quickstart. This request searches the web and the news for three results each.
You can try Search without an API key, up to 10 calls a minute and 50 calls a day per IP address. Reading the page text of each result needs an API key. See Rate limits.

Read the response

data holds one list for each source you asked for. Every row has a position. This real response is cut to the first rows of each list. Each news row also has an imageUrl.
A web row has a url, a title and a description. date appears when the date of the page is known, and category is github, research or pdf when the page is one of those. A news row has a snippet instead of a description. costUsd is what the call cost. If a source answers with fewer results than you asked for, or fails while the others work, the call still succeeds and warning says which. If every source fails, the call fails and costs nothing.

Options

These fields belong to this page. The full field list is in the API reference.

Price

$1 per 1,000 searches of up to 10 results: $0.001 for each 10 results from each source, and for each list of papers or videos. A search that returns nothing costs nothing. Page text adds $1 per 1,000 results read, and `summary` or `json` with a schema or a prompt add $3 per 1,000 more each. See Pricing.

Choose another API

  • Scrape reads one page you already know.
  • Extract pulls the fields you name from pages, each checked against the page.
  • Agent searches and reads for you until your fields are filled.
  • Crawl reads every page of a site in the background.
  • Map lists the pages of a site without reading them.