Web search

Find the right page first.

Discover relevant public pages with one cacheable GET request. Results contain ordered titles, URLs, and concise snippets ready for an agent to evaluate.

Request

curl --get 'https://webstractor.com/api/search' \
  --data-urlencode 'q=Cloudflare Workers documentation' \
  --data-urlencode 'format=json'

Parameter	Required	Description
`q`	Yes	A search query of up to 200 characters.
`limit`	No	An integer from `1` to `10`. Default: `10`.
`language`	No	Canonical BCP 47 language tag. Default: `en-US`.
`country`	No	Two-letter ISO country code. Default: `US`.
`site`	No	A hostname such as `github.com` or `linkedin.com`. Results are restricted to that hostname and its subdomains; do not include a scheme or path.
`format`	No	`json` (default) or `markdown`.

JSON response

JSON uses schema version 1 with type: "feed" and source: "web-search". The ordered items are content-free document entities. Request Markdown for readable snippets, then fetch a selected url with /api/extract when full page content is needed.

{
  "schemaVersion": 1,
  "type": "feed",
  "source": "web-search",
  "id": "web-search:sha256:…",
  "url": "https://webstractor.com/api/search?q=Cloudflare+Workers+documentation&format=json",
  "title": "Search results for Cloudflare Workers documentation",
  "author": null,
  "publishedAt": null,
  "media": [],
  "attributes": {
    "feedType": "web-search",
    "query": "Cloudflare Workers documentation",
    "language": "en-US",
    "country": "US",
    "resultCount": 1
  },
  "items": [
    {
      "type": "document",
      "source": "web-search",
      "id": null,
      "url": "https://developers.cloudflare.com/workers/",
      "title": "Cloudflare Workers docs",
      "author": null,
      "publishedAt": null,
      "media": [],
      "attributes": {}
    }
  ]
}

Behavior

Results are ordered by estimated relevance. Their order is their rank.
Successful searches are cached for up to one hour. Identical cache hits are not billed and bypass search work.
Search uses the standard 60 requests per client per minute limit.
Search snippets are discovery metadata, not full page content.
Language and country controls affect result selection. Safe search remains strict and cannot be disabled.
Use site=linkedin.com to discover indexed public profile and company URLs. Results are search metadata and source links, not native LinkedIn records or full page extraction.
Result availability, freshness, and ranking can change. Pagination is not currently supported.

Agent workflow

Call /api/search to discover candidate URLs.
Select only the relevant results.
Call /api/extract for full Markdown or typed page data.