subimpact WhatsApp us

RAG Web Browser Lite: Markdown for AI Agents

RAG Web Browser Lite takes a search query or a single URL, pulls the top N Google organic results, fetches those pages over raw HTTP and returns clean Markdown, plain text or cleaned HTML per page, with page metadata, OpenGraph and JSON-LD, the search result it came from and HTTP status. It runs without a browser and costs $1.00 per 1,000 extracted pages. Run it from Apify Console, the API or the standby endpoint.

from $0.001 per page extracted, $1.00 per 1,000 · Updated Nov 2026, version 0.1.16

Key facts

Price
from $0.001 per page extracted, $1.00 per 1,000
Pricing model
Pay per event
Output
JSON, CSV, Excel, XML, RSS via Apify datasets
Access
Apify Console, REST API, Python and JS clients, MCP tool web_page_markdown
Version
Updated Nov 2026, version 0.1.16
Actor
subimpact/rag-web-browser-lite

Who it is for

  • AI engineers wiring a search and fetch tool into an agent
  • RAG teams building a corpus with clean Markdown instead of HTML
  • Developers replacing a heavier browser-based fetcher to cut compute
  • Analysts who want the top Google results of a query as readable text

Use cases

Give an agent a search tool

Send query and maxResults, for example best crm for startups with 3 results. Each row pairs a searchResult object of title, description, url and position with the extracted markdown of that page, so an agent gets both the ranking context and the page body in one response, with no second call to open each link.

Fetch a specific page only

Pass a URL as the query, such as https://www.cnn.com, and maxResults is ignored, so the run fetches that page alone. metadata returns the title, description, languageCode, canonicalUrl, openGraph and jsonLd, while crawl holds httpStatusCode and requestStatus, which tells you whether content was actually extracted or the fetch failed.

Strip navigation before conversion

removeElementsCssSelector defaults to nav, footer and script elements. Swap in your own selector when a site wraps content in a noisy container, or set a selector that matches nothing, such as dummy_keep_everything, to keep the page intact. Cleaner input means fewer tokens spent on menus and cookie banners in your context window.

Serve it as a standby endpoint

The actor can run in standby mode, giving a single HTTP endpoint that takes query and maxResults as parameters and returns the extracted pages. That suits an agent runtime or a workflow tool that needs a plain GET rather than a run-and-poll cycle, and the token is passed as a query parameter.

Localise a search

Set gl to my, sg or uk and hl to the interface language, such as en, ms or zh, to change which Google results are returned for the same query. serpProxyGroup and serpMaxRetries control how the search is fetched, and proxyConfiguration covers the target page fetches, so market-specific research uses the right country and language.

Output fields

FieldTypeDescription
crawl object/null HTTP result metadata for the page: httpStatusCode, httpStatusMessage, loadedAt, uniqueKey, requestStatus ("handled" = content extracted; "failed" = fetch failed, no content).
metadata object/null Extracted page metadata: url, title, description, languageCode, canonicalUrl, openGraph, jsonLd, headers.
searchResult object/null The organic search result this page came from: title, description (snippet), url, resultType, position. null for direct-URL requests.
query string/null The search term this row's page was found for. null for direct-URL requests.
markdown string/null Page content converted to Markdown (when 'markdown' was selected in outputFormats).
text string/null Page content as plain text (when 'text' was selected in outputFormats).
html string/null Cleaned body HTML (when 'html' was selected in outputFormats).

Sample output

From one of our own test runs on 20 Nov 2026, trimmed. Personal data is redacted.

1 rows, first 3 shown

[
  {
    "crawl": {
      "httpStatusCode": 200,
      "httpStatusMessage": "OK",
      "loadedAt": "2026-11-20T02:41:49.288Z",
      "uniqueKey": "https://example.com/",
      "requestStatus": "handled"
    },
    "metadata": {
      "url": "https://example.com/",
      "title": "Example Domain",
      "languageCode": "en",
      "headers": {
        "content-type": "text/html; charset=utf-8",
        "last-modified": "Thu, 19 Nov 2026 22:10:26 GMT"
      }
    },
    "searchResult": null,
    "query": null,
    "markdown": "Example Domain\n\nThis domain is for use in documentation examples without needing permission. This is not a service; avoid relying on it for testing and monitoring purposes."
  }
]

Input

FieldTypeRequiredDescriptionDefault
query string No Enter Google Search keywords or a URL of a specific web page. Examples: - san francisco weather - https://www.cnn.com - function calling site:openai.com https://example.com
maxResults integer No The maximum number of top organic Google Search results whose web pages will be extracted. If `query` is a URL, this field is ignored and the Actor only fetches the specific web page. 3
outputFormats array No Select one or more formats to which the target web pages will be extracted and saved in the resulting dataset. ["markdown"]
gl string No Two-letter country code for Google Search, e.g. us, my, uk, sg. us
hl string No Two-letter language code for Google Search, e.g. en, ms, zh. en
serpProxyGroup string No Apify Proxy group used for fetching Google Search results. GOOGLE_SERP
serpMaxRetries integer No The maximum number of times the Actor will retry fetching the Google Search results on error. 2
proxyConfiguration object No Apify Proxy configuration used for scraping the target web pages. {"useApifyProxy":true}
removeElementsCssSelector string No A CSS selector matching HTML elements that will be removed from the DOM before converting to text, Markdown, or saving as HTML. Set to a non-existent selector like `dummy_keep_everything` to disable removal. nav, footer, script, style, noscript, svg, img[src^='data:'], [role="alert"], [role="banner"], [role="dialog"], [role="alertdialog"], [role="region"][aria-label*="skip" i], [aria-modal="true"]

Example input

{
  "query": "web browser for RAG pipelines -site:reddit.com",
  "proxyConfiguration": {
    "useApifyProxy": true
  },
  "removeElementsCssSelector": "nav, footer, script, style, noscript, svg, img[src^='data:'],\n[role=\"alert\"],\n[role=\"banner\"],\n[role=\"dialog\"],\n[role=\"alertdialog\"],\n[role=\"region\"][aria-label*=\"skip\" i],\n[aria-modal=\"true\"]"
}

Pricing

EventWhat it meansPriceLowest tier
page-extracted One web page successfully extracted to Markdown/text/HTML $0.001 $0.001

$1.00 for 1,000 page extracted units at the listed price.

Prices from the Apify Store on 20 November 2026. Apify shows the final price before each run.

How it compares with other Apify actors

Price per 1,000 pages at Apify Store list prices. Every row is a live listing on the Store, read on 20 Nov 2026, so a buyer can check each number at the source.

Actor Price per 1,000 pages (Free plan) Higher plans (Gold) Start fee per run Platform usage Users Rating Last build
RAG Web Browser Lite — Cheap Markdown for AI Agents This API Pay per event $1.00 $1.00 None Paid separately 1 No reviews yet 18 Nov 2026
RAG Web Browser Direct match, pay per event $2.28 $1.33 $0.000225 Included 213,874 4.7 (35 reviews) 16 Nov 2026
RAG Web Browser — URL & Website to Markdown for LLM & RAG Direct match, pay per event $2.00 $2.00 None Included 13 No reviews yet 18 Nov 2026
RAG Web Browser Direct match, pay per event $12.00 $8.00 $0.005 Included 20 No reviews yet 20 Oct 2026
RAG Web Browser Scraper Direct match, pay per event $2.00 $2.00 $0.00005 Paid separately 12 No reviews yet 11 Nov 2026
Where RAG Web Browser is stronger

Where RAG Web Browser is stronger

  • Much larger user base and the reference actor in this lane
  • Full browser rendering, so pages that need scripts still load
  • Output shape that other RAG pipelines are already written against

Where this API is stronger

  • Raw HTTP with no browser, so the platform usage on your account stays small
  • Same output shape as theirs, so existing pipelines need no reshaping
  • Light container per page, which keeps run overhead down

Full comparison: RAG Web Browser vs this API

Where RAG Web Browser — URL & Website to Markdown for LLM & RAG is stronger

Where RAG Web Browser — URL & Website to Markdown for LLM & RAG is stronger

  • Crawls links with depth limits, same domain and URL pattern filters
  • Markdown, plain text or HTML plus page metadata as output options
  • Reports providers tried and fetch timing per page

Where this API is stronger

  • Google search and page fetch in one small actor
  • Output keys crawl, searchResult, metadata and markdown that fit existing RAG code
  • No browser step, so the compute you pay for stays small
Where RAG Web Browser is stronger

Where RAG Web Browser is stronger

  • Optional link extraction and HTML output as add-ons
  • Reading time, author, site name and image URL per page
  • Search rank recorded for rows that came from a query

Where this API is stronger

  • No start fee before the first page is extracted
  • Markdown, text and HTML returned together in one row
  • Raw HTTP fetch keeps platform usage low
Where RAG Web Browser Scraper is stronger

Where RAG Web Browser Scraper is stronger

  • Proxy group selection and several queries in one input
  • Language and HTTP status metadata per page

Where this API is stronger

  • Google domain and language are explicit search inputs
  • No start fee before the first page
  • CSS selector removal to strip page chrome before the Markdown is written

This lane is crowded and the Apify listing is the reference implementation, with browser rendering as its main advantage. The s-r, parseforge and datapilot actors convert pages with differing metadata. Ours matches the Apify output shape but fetches over raw HTTP, so search and page extraction stay cheap while existing pipelines keep working unchanged.

Prices are Apify Store list prices, checked 20 Nov 2026. Platform usage 'Paid separately' means Apify bills compute on top of the event price. Check each listing before you buy.

How to run it

  1. Enter Google keywords such as best crm for startups, or a single URL such as https://www.cnn.com, in the query field.
  2. Set maxResults, for example 3, which caps how many top organic results are fetched. It is ignored when query is a URL.
  3. Choose outputFormats from markdown, text and html. markdown is the default and the usual choice for an LLM.
  4. Optionally set gl and hl for country and language, and adjust removeElementsCssSelector if a page converts badly.
  5. Run the actor, then read markdown, metadata and searchResult from the dataset, or call the standby endpoint for a single request.

cURL

curl -X POST "https://api.apify.com/v2/acts/subimpact~rag-web-browser-lite/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"web browser for RAG pipelines -site:reddit.com","proxyConfiguration":{"useApifyProxy":true},"removeElementsCssSelector":"nav, footer, script, style, noscript, svg, img[src^='data:'],\n[role=\"alert\"],\n[role=\"banner\"],\n[role=\"dialog\"],\n[role=\"alertdialog\"],\n[role=\"region\"][aria-label*=\"skip\" i],\n[aria-modal=\"true\"]"}'

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("subimpact/rag-web-browser-lite").call(run_input={"query":"web browser for RAG pipelines -site:reddit.com","proxyConfiguration":{"useApifyProxy":true},"removeElementsCssSelector":"nav, footer, script, style, noscript, svg, img[src^='data:'],\n[role=\"alert\"],\n[role=\"banner\"],\n[role=\"dialog\"],\n[role=\"alertdialog\"],\n[role=\"region\"][aria-label*=\"skip\" i],\n[aria-modal=\"true\"]"})

for item in client.dataset(run["default_dataset_id"]).iterate_items():
    print(item)

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('subimpact/rag-web-browser-lite').call({"query":"web browser for RAG pipelines -site:reddit.com","proxyConfiguration":{"useApifyProxy":true},"removeElementsCssSelector":"nav, footer, script, style, noscript, svg, img[src^='data:'],\n[role=\"alert\"],\n[role=\"banner\"],\n[role=\"dialog\"],\n[role=\"alertdialog\"],\n[role=\"region\"][aria-label*=\"skip\" i],\n[aria-modal=\"true\"]"});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

console.log(items);

MCP

claude mcp add subimpact-data -e APIFY_TOKEN=<YOUR_APIFY_TOKEN> -- npx -y @subimpact/data-mcp

# Then ask your agent to use the web_page_markdown tool.

n8n

Use the Apify node (Run an Actor).
Actor ID: subimpact/rag-web-browser-lite
Input JSON: paste the example input above.

MCP users: @subimpact/data-mcp exposes these actors as tools, so an agent can call web_page_markdown directly.

Limits

  • No browser, so JavaScript-heavy pages can return stubs that are not charged.
  • Pages with under 250 characters of real content are treated as empty.
  • SERP fetching is billed as platform usage rather than as an event.
  • Only the top organic Google results are fetched, with no deep paging.

Questions people ask

What does a page cost?

Each extracted page costs $0.001, which is $1.00 per 1,000 pages. Standard Apify platform usage is billed separately, and the raw HTTP engine keeps that low because no browser starts. Pages that fail, are blocked or return under 250 characters of real content are not charged.

Does it use a browser?

No. Pages are fetched over raw HTTP and converted with cheerio and turndown in a small container, which is why compute stays low. The trade-off is that JavaScript-heavy sites, including Reddit, can return stub pages, and those are caught by the content gate rather than billed.

Is the output shape the same as apify/rag-web-browser?

Yes. Rows carry crawl, searchResult, metadata, markdown, and text or html when those formats are selected, so an existing pipeline can switch to this actor without reshaping records. markdown is the default output format.

How many search results can it process?

maxResults takes 1 to 100 and defaults to 3, and each result becomes its own row with its markdown. When the query is a URL instead of a search term, maxResults is ignored and only that page is fetched.

Can I use it as an HTTP endpoint instead of a run?

Yes, in standby mode. A single GET with query and maxResults plus your token returns the extracted pages without creating a run and polling for it, which suits agent runtimes and workflow tools that expect one call.

Why did a page come back empty or failed?

crawl.requestStatus reports handled when content was extracted and failed when the fetch did not produce usable content, with httpStatusCode alongside. Pages under 250 characters of real content are treated as empty and are not charged, so an empty row is a signal rather than a cost.

How do I make results country specific?

Set gl to the country code, for example us, my, sg or uk, and hl to the language code such as en, ms or zh. That changes the Google results returned for the query, so the same search can surface different publishers per market.

Need a data source that is not here?

We build and run scrapers for sites that have no public API. Tell us the site, the fields and how often you need them.