RAG Web Browser Lite — Cheap Markdown for AI Agents

A lean, drop-in alternative to apify/rag-web-browser for AI agents and RAG pipelines.

Same output shape, fixed event price. Queries Google Search, scrapes the top N pages, and returns clean Markdown — but over raw HTTP only (no Playwright browser), so each page is billed as a flat $0.001 pay-per-event ($1.00 per 1,000 pages; standard Apify platform usage — compute and proxy traffic — applies separately and is paid by the buyer, as with any Apify actor; the raw-HTTP engine keeps that usage small).

Why it's cheaper

	apify/rag-web-browser	rag-web-browser-lite
Engine	Crawlee + Playwright + MCP SDK	raw HTTP + cheerio + turndown
Browser launch per page	Yes (heavy)	Never
Billing	PPE: $2.28 per 1,000 fetches, $3.48 per 1,000 searches (Free plan, 24 Oct 2026)	PPE event: $0.001/page + platform usage paid by buyer
Container	~1GB+ (Playwright)	~200MB
Output shape	{crawl, searchResult, metadata, markdown}	Identical

The official actor's own docs say raw HTTP is "about two times faster" than the browser tool — we made that the only mode.

Usage

Normal mode (one request per run)

curl -X POST "https://api.apify.com/v2/acts/subimpact~rag-web-browser-lite/runs" \
  -H "Content-Type: application/json" \
  -d '{"query": "best crm for startups", "maxResults": 3}'

Standby mode (agent-facing HTTP endpoint)

$curl "https://subimpact--rag-web-browser-lite.apify.actor/search?query=apify+mcp&maxResults=2&token=<APIFY_API_TOKEN>"

Input

Field	Type	Default	Description
query	string	https://example.com	Search term or URL; if omitted entirely, this default URL is used
maxResults	int	3	Top N organic results to extract (1-100)
outputFormats	array	["markdown"]	markdown, text, html
gl / hl	string	us / en	Google country / language
serpProxyGroup	string	GOOGLE_SERP	Proxy group for SERP fetch
serpMaxRetries	int	2	SERP retries (0-5)
proxyConfiguration	object	Apify proxy	Proxy for target pages
removeElementsCssSelector	string	nav/footer/script/...	Elements to strip before conversion

Output row

{
  "crawl": { "httpStatusCode": 200, "httpStatusMessage": "OK", "loadedAt": "...", "uniqueKey": "...", "requestStatus": "handled" },
  "searchResult": { "title": "...", "description": "...", "url": "...", "resultType": "ORGANIC", "position": 1 },
  "metadata": { "url": "...", "title": "...", "description": "...", "languageCode": "en", "canonicalUrl": "...", "openGraph": [...], "jsonLd": [...], "headers": {...} },
  "markdown": "# ..."
}

Pricing

• $0.001 per page extracted (pay-per-event, page-extracted, $1.00 per 1,000 pages)
• Standard Apify platform usage (compute, proxy, ingress) is paid by the buyer; the raw-HTTP engine keeps it low
• No charge for failed/blocked/empty pages — only pages with ≥250 chars of real content are billed
• SERP fetch is billed as platform usage (GOOGLE_SERP proxy), not as an event

How this compares

The output shape (crawl, searchResult, metadata, markdown) is the same as apify/rag-web-browser, so it drops into an existing pipeline without reshaping. No browser is launched: pages are fetched over raw HTTP with cheerio and turndown in a ~200MB container instead of a ~1GB Playwright image, which keeps the platform usage you pay for small. At $1.00 per 1,000 pages it is below every alternative in this lane on the Store: apify/rag-web-browser lists $2.28 per 1,000 fetches on the Free plan (and $3.48 per 1,000 searches).

Actor	Price per 1,000 pages	Platform usage
subimpact/rag-web-browser-lite	$1.00	paid by you
apify/rag-web-browser	$2.28 (fetch), $3.48 (search)	see listing
s-r/rag-web-browser	$2.00	see listing
parseforge/rag-web-browser	$12.00	see listing
datapilot/rag-web-browser-scraper	$2.00	see listing

Prices are Apify Store list prices on the Free plan, checked 24 October 2026. Competitor prices can change; check their listings.

Development

npm install
npm test        # unit tests (parser + content pipeline)
apify push      # build to Apify

Notes

• SERP parsing reuses the battle-tested goto-map + organic parser from google-serp-ai-overviews (verified 2026-08-27: top-organic 100% mappable).
• GOOGLE_SERP proxy requires http:// (not https://) for the SERP fetch.
• Standby detection: Actor.getEnv().metaOrigin === 'STANDBY'; port from Actor.config.get('containerPort').
• Reddit and other JS-heavy sites that block raw HTTP return stub pages — those are not charged (content gate).