Web Search API for AI Agents

Web Search Scraper runs web searches without an API key on Brave Search, DuckDuckGo and Bing and returns clean JSON: position, title, URL, domain, site name, snippet and publication date. Turn on scrapeContent and it also opens the top results and returns their clean main text in markdown (no menus, ads or footers), ready for an LLM, an AI agent tool call or a RAG pipeline.

US$ 1.00 per 1,000 result pages (Brave pages carry up to 20 results). Page text US$ 0.40 per 1,000 pages. No monthly fee, no start fee. Error rows are free.

At a glance

• Price per 1,000 result pages: US$ 1.00 (down to US$ 0.85 on Gold and above), no run start fee
• Price per 1,000 pages read (optional): US$ 0.40, only pages that gave text
• Engines: Brave Search (20 results per page), DuckDuckGo, Bing; one or several merged, duplicates removed
• Fallback: If the chosen engine is blocked or empty, Brave and DuckDuckGo take over automatically
• Freshness: Past day, week, month or year; publication date on each result when the engine has it
• Page text: Main text in markdown or plain text, with title, author, date and length
• Error rows: Free, with an errorCode

One real row

Query how to lower cholesterol naturally, engines Brave + DuckDuckGo, scrapeContent on (collected on 2026-09-30, content cut here):

{
  "type": "result",
  "query": "how to lower cholesterol naturally",
  "engine": "brave",
  "position": 1,
  "engineRank": 1,
  "page": 1,
  "title": "Top 5 lifestyle changes to improve your cholesterol - Mayo Clinic",
  "url": "https://www.mayoclinic.org/diseases-conditions/high-blood-cholesterol/in-depth/reduce-cholesterol/art-20045935",
  "domain": "mayoclinic.org",
  "siteName": "Mayo Clinic",
  "snippet": "Soluble fiber helps block some cholesterol from being absorbed into your bloodstream. You can find soluble fiber in foods such as oatmeal, kidney beans, Brussels sprouts, apples and pears...",
  "content": "High cholesterol raises your risk of heart disease and heart attacks. Medicines can help improve your cholesterol. But if you'd rather start with some healthy lifestyle changes...\n\n## 1. Eat heart-healthy foods\n\n...",
  "contentLength": 5143,
  "contentTruncated": false,
  "contentTitle": "Top 5 lifestyle changes to improve your cholesterol",
  "contentAuthor": "Mayo Clinic Staff",
  "contentDate": "2026-01-21",
  "contentError": null
}

What you can do with it

• Give an AI agent a search tool: call the Actor with a query (or through the Apify MCP server) and feed the markdown straight into the prompt. One call returns both the links and the text.
• RAG and research pipelines: collect the top pages for hundreds of questions, with their text, title, author and date, in one dataset.
• SEO and rank tracking: positions, domains and snippets for your keywords on Brave, DuckDuckGo and Bing, by country and language.
• News and freshness monitoring: results from the past day or week, with dates.
• Lead and market research: which domains show up for a topic, with site names.

Input

Field	What it does
queries	One query per line. Operators like site:, quotes and -word work as on the engine
engines	brave (default), duckduckgo, bing; several are merged rank by rank
maxResults	Default 10, up to 200 per query
fallbackEngines	Default on: if the chosen engines return nothing, try Brave, then DuckDuckGo
countryCode, languageCode	Region and language, like us + en, de + de, br + pt
timeRange	any, day, week, month, year
scrapeContent	Also open the top results and return their main text
maxContentPages	How many top results to read per query (default 5)
contentFormat	markdown (default) or text
maxContentChars	Cut long texts here (default 30,000); contentLength keeps the full size

{
  "queries": ["what is retrieval augmented generation", "site:docs.python.org asyncio"],
  "engines": ["brave", "duckduckgo"],
  "maxResults": 10,
  "timeRange": "year",
  "scrapeContent": true,
  "maxContentPages": 5
}

Output fields

query, engine, position (merged rank), engineRank, page, title, url, domain, siteName, snippet, date (ISO, from Brave and DuckDuckGo when available), dateText (as shown, like "2 days ago"), and with scrapeContent: content, contentLength, contentTruncated, contentTitle, contentAuthor, contentDate, contentLanguage, finalUrl (after redirects), contentError (why a page gave no text; not charged), scrapedAt.

Error codes (never charged)

errorCode	Meaning
NO_RESULTS	The engines returned no result for this query
BLOCKED	The engines did not answer after retries and the fallback; run again
INVALID_INPUT	A value the Actor does not understand
NOT_REACHED, UNEXPECTED	The run timeout arrived, or something unforeseen broke; only that query is lost

Pricing

Event	Per 1,000
Search results page (with at least one result)	US$ 1.00
Page content extracted (only with scrapeContent)	US$ 0.40

Ten results from Brave are one page: 1,000 searches of 10 results cost US$ 1.00. With two engines each query uses one page per engine. Adding the text of the top 5 pages costs up to US$ 2.00 more per 1,000 queries. Blocked or empty pages, pages without text and error rows are free, and there is no Apify compute bill on top.

Limitations

• No Google. Google now requires JavaScript for every search, and the ways around it cost more than this whole price. Brave Search runs its own index; DuckDuckGo draws largely on Bing's.
• Bing sometimes answers anonymous visitors with loosely related results, so it is never used as a fallback; pick it explicitly if you want it.
• Pages that need a login, are built only with JavaScript or refuse automated visits come back with a contentError and are not charged. PDFs are skipped.
• Only public search results and public pages. This Actor respects the sites' terms and does not solve captchas.

FAQ

Do I need an API key? No. Searches run on Brave Search, DuckDuckGo and Bing without an API key, through the Apify proxy.

What does a run cost? US$ 1.00 per 1,000 result pages with at least one result (a Brave page carries up to 20 results), plus US$ 0.40 per 1,000 pages read when scrapeContent is on. 1,000 searches of 10 results on Brave cost US$ 1.00. Empty or blocked pages, pages without text and error rows are free, and there is no Apify compute bill on top.

Does it search Google? No. Google now requires JavaScript for every search. Use engines to pick brave (default), duckduckgo, bing or several merged.

How do I get the text of the pages, not just the links? Turn on scrapeContent. The Actor opens the top maxContentPages results per query (default 5) and returns their main text in content, as markdown or text (contentFormat), cut at maxContentChars.

Can an AI agent use it as a search tool? Yes. Call the Actor with a query, or through the Apify MCP server, and feed the markdown in content straight into the prompt.

How many results can I get per query? Up to 200 with maxResults (default 10). Use timeRange (day, week, month, year) for fresh results, and countryCode with languageCode for a region.

Changelog

• 2026-10-17: FAQ in this page.
• 2026-09-30, 0.1: first release. Brave, DuckDuckGo and Bing, merged results with automatic fallback, country, language and time filters, optional page text in markdown for LLMs, free error rows.