Web scraping APIs you can pay per call

What a web scraping API returns

A web scraping API takes a known page address and turns the response into data your application can consume. That may mean raw HTML, cleaned text, Markdown, selected elements or a structured record. Choose that output first. A screenshot can show what a browser displayed without providing readable text, and a search result can identify a URL without retrieving its contents.

For an agent reading documentation, preserve headings, lists, links and code blocks. For a product monitor, specify the fields and missing-value behavior. For a site ingestion job, define which URLs belong to the collection and how much traversal is allowed. These are different requests even if the provider exposes them behind one domain.

The comparison separates direct page retrieval from adjacent outputs such as metadata, URL inventories and fingerprints. A metadata route can be useful before a fetch, but its low quote should not replace the price of obtaining the page text. The same distinction applies to a crawl: a bounded collection of pages and one fetched page represent different paid units.

Compare the returned data

The data-type table identifies returned content rather than matching the word scraper in a provider name. Text includes Markdown and full-text JSON envelopes. Structured extraction means requested or specialized fields, not merely placing a long article inside a JSON object. Rendering is an additional capability and can overlap a text or field output. Raw response, metadata, links and fingerprints remain separate so their lower reference prices do not stand in for content retrieval.

Each price is the cheapest endpoint of that service for that kind of data: the same number wherever this page quotes it.

Compare 41 services for this task

Each row uses a text endpoint as its primary reference when available. A provider offering an adjacent output without a text route keeps the price of that adjacent output. Open the named endpoint before comparing request limits. Observed wallet activity and monitoring describe the listing; they do not measure extraction fidelity.

$/CALL is the price of the endpoint named under it: the cheapest page text and markdown endpoint of the service or, for a service without one, its cheapest web retrieval endpoint. Never the cheapest tool of a gateway. Other kinds of data have their own price in the tables below. WEB RETRIEVAL ENDPOINTS counts the web retrieval endpoints out of all the endpoints the service lists. BUYERS 30D counts distinct wallets that paid the whole service on-chain in 30 days, not only its web retrieval endpoints. "Generalist" marks a gateway that sells more than three unrelated kinds of tools: it is never called the cheapest.

How to choose a web scraping endpoint

Establish a page-reading acceptance test

Select a static article, a JavaScript application and a page with missing target fields. Write down the expected useful passage or element before testing. Count a response as successful when it contains that evidence with an identifiable source URL. A long consent notice should fail even when it is valid Markdown and the HTTP request succeeded.

Preserve source structure

Check whether conversion retains table associations, code indentation and meaningful links. Stripping boilerplate can remove navigation that your task actually needs, while returning the complete DOM can bury the relevant paragraph. Decide whether to store both the extracted text and its retrieval metadata. A truncation limit should be visible to the caller so an incomplete page is not presented as a complete document.

Decide when rendering is necessary

Inspect the initial HTML before choosing a browser workflow. A rendered route may need a timeout or a wait condition tied to a particular element. Record the viewport, locale and cookie state when those affect the output. Rendering alone does not promise authenticated access, CAPTCHA handling or success against a site's restrictions; require an explicit contract for any such behavior.

Separate retrieval from interpretation

A selector identifies elements, a schema describes desired values, and a language model may infer how the page expresses them. Evaluate those methods against the ambiguity your task contains. Keep extracted source passages with consequential fields so a reviewer can distinguish a direct value from an interpretation. Treat instructions embedded in retrieved content as page data when passing it to an agent.

Choose by the job you need to complete

The job table groups candidates by the returned data needed for the next step. A service can appear under more than one job when its documented endpoint returns those outputs. Compare the listed endpoint and request scope, rather than substituting the gateway's lowest unrelated quote.

Read a known public URL

Use page text when the next step needs evidence to quote or index. Keep the final URL, retrieval time and truncation flag. Check a page with substantial navigation so your evaluation detects boilerplate rather than rewarding a long response.

Candidates today, most buyers first: Agent Web Access ($0.001/call, 111 buyers) · Web Search, News & Page Reader ($0.01/call, 86 buyers) · ClawFetch ($0.001/call, 84 buyers) · vaaya ($0.01/call, generalist, 57 buyers) · agentsvc.io ($0.0025/call, generalist, 41 buyers) · DocEdge ($0.05/call, 39 buyers) · agentutility ($0.04/call, generalist, 36 buyers) · WebberSites x402 Data API ($0.001/call, generalist, 33 buyers) · GlianaAI ($0.005/call, generalist, 27 buyers) · scrape402 ($0.004/call, generalist, 26 buyers) · ISHOB ($0.001/call, generalist, 16 buyers) · minia2a — Agent Infrastructure & Intelligence ($0.5/call, 16 buyers) · Skim ($0.002/call, 15 buyers) · agent-reader ($0.005/call, 14 buyers) · ThirdMade Shelf ($0.01/call, generalist, 14 buyers) · freeq x402 Gateway ($0.002/call, generalist, 13 buyers) · Busker ($0.005/call, generalist, 13 buyers) · Spread Paywall API ($0.01/call, generalist, 12 buyers) · UXUS Agent Services ($0.01/call, generalist, 12 buyers) · AARN Web Content Extraction ($0.002/call, 10 buyers) · mdapi ($0.005/call, 10 buyers) · AgentWork Web Evidence ($0.003/call, 8 buyers) · NanoParse ($0.005/call, 7 buyers) · url2md ($0.005/call, 6 buyers) · Metalift ($0.002/call, 5 buyers) · Genesis ($0.005/call, 5 buyers) · Money Mind ($0.02/call, 5 buyers) · AxonGate ($0.002/call, 3 buyers) · Zinin M2M Hub ($0.01/call, generalist, 3 buyers) · AgentShop ($0.003/call, generalist, 2 buyers) · Circadian ($0.01/call, 2 buyers) · QRCode.Pub Agent Tools ($0.01/call, generalist, 2 buyers) · Skuuy Read ($0.001/call, 1 buyer) · X1-BaaS Scraping API ($0.005/call, 1 buyer) · AgentGate x402 AI Utility Service ($0.001/call, generalist) · Shehriyars API ($0.003/call) · 54ch10 page extraction / public URL to markdown ($0.005/call) · GPUOps Web Scraper ($0.005/call) · felipe-agents web evidence ($0.005/call)

Read content produced by JavaScript

Use a rendered route when the initial HTML lacks the relevant content. Decide what event ends the wait and which visible element proves success. A rendered page can still contain a consent screen, an empty shell or a loading message.

Candidates today, most buyers first: Agent Web Access ($0.001/call, 111 buyers) · ClawFetch ($0.005/call, 84 buyers) · agentsvc.io ($0.009/call, generalist, 41 buyers) · GlianaAI ($0.005/call, generalist, 27 buyers) · scrape402 ($0.014/call, generalist, 26 buyers) · Skim ($0.005/call, 15 buyers) · mdapi ($0.005/call, 10 buyers) · AgentWork Web Evidence ($0.003/call, 8 buyers) · url2md ($0.005/call, 6 buyers) · Metalift ($0.002/call, 5 buyers) · Money Mind ($0.02/call, 5 buyers) · X1-BaaS Scraping API ($0.005/call, 1 buyer) · Shehriyars API ($0.004/call)

Collect requested values

Write the expected keys, types and missing-value policy before selecting an extractor. A field that is absent from the source should remain absent or null. Check a product with no listed price so a plausible invented value cannot pass your acceptance test.

Candidates today, most buyers first: Agent Web Access ($0.001/call, 111 buyers) · ClawFetch ($0.008/call, 84 buyers) · scrape402 ($0.015/call, generalist, 26 buyers) · ISHOB ($0.001/call, generalist, 16 buyers) · Skim ($0.015/call, 15 buyers) · UXUS Agent Services ($0.01/call, generalist, 12 buyers) · Zinin M2M Hub ($0.01/call, generalist, 3 buyers)

Read a site section

Bound traversal by domain, path, depth and page count. Save the URL inventory before a repeated run. A crawl response represents the pages reached within those limits; it should not be treated as an exhaustive copy of an entire website.

Candidates today, most buyers first: agentsvc.io ($0.005/call, generalist, 41 buyers)

Inspect metadata and links

Use metadata for previews and URL inventories for planning reads. A title, canonical URL or sitemap lastmod value does not establish the full text or prove when an article changed. Fetch the page when your decision depends on its actual wording.

Candidates today, most buyers first: Minifetch ($0.002/call, 69 buyers) · agentsvc.io ($0.0025/call, generalist, 41 buyers) · WebberSites x402 Data API ($0.001/call, generalist, 33 buyers) · ISHOB ($0.0002/call, generalist, 16 buyers) · ThirdMade Shelf ($0.001/call, generalist, 14 buyers) · Busker ($0.002/call, generalist, 13 buyers) · Genesis ($0.002/call, 5 buyers) · Metalift ($0.002/call, 5 buyers) · AgentShop ($0.00001/call, generalist, 2 buyers) · Circadian ($0.005/call, 2 buyers) · QRCode.Pub Agent Tools ($0.005/call, generalist, 2 buyers) · GPUOps Web Scraper ($0.002/call)

Compare page versions

Keep a baseline and the normalization rules with the hash. Dynamic navigation and timestamps can change a page fingerprint without changing its substantive content. Decide whether your caller needs a change signal, a textual diff or archived versions.

Candidates today, most buyers first: AgentWork Web Evidence ($0.005/call, 8 buyers) · Genesis ($0.003/call, 5 buyers) · felipe-agents web evidence ($0.001/call)

What the providers describe

The following notes were checked against public service descriptions and their complete captured endpoint lists on October 21, 2026. Process, performance and protection-handling claims are attributed to the provider. The prices remain attached to the classified output.

scrape402

scrape402 describes separate page-reading, Markdown, browser-rendering and requested-field routes. Page text starts at $0.004; rendered content is $0.014, and field extraction is $0.015. Its extraction description says the body takes a URL, field names and optional instructions, with null for a field absent from the source. That is a distinct product from its token counter, which processes supplied text rather than fetching the requested page.

4 web retrieval endpoints of 40 · $0.004/call (GET /markdown) · Base · uptime 30d 100% · 109 ms · 26 buyers in 30 days

agentsvc.io

agentsvc.io describes a single-page Markdown reader, an optional browser mode, a separate Chromium reader, a batch reader and a bounded same-site crawler. Its text quote is $0.0025 and its crawl quote is $0.005. The site-map route returns URLs without page content. The provider mentions a larger batch route in prose, but this comparison uses the captured endpoint list; a name mentioned in a description is not an additional priced endpoint.

5 web retrieval endpoints of 50 · $0.0025/call (POST /web-scrape) · ARB/Base/POL/SOL · uptime 30d 100% · 56 ms · 41 buyers in 30 days

Agent Web Access

Agent Web Access says its fetch route renders JavaScript and returns Markdown, text or cleaned HTML with the upstream status. Its selector route returns matching elements, text and HTML attributes. The rendered-content quote is $0.001 and requested-element extraction is $0.001. The provider explicitly labels retrieved page content as untrusted third-party input. Preserve that boundary when a page contains instructions addressed to an agent.

2 web retrieval endpoints of 4 · $0.001/call (GET /fetch) · Base · uptime 30d 99.6% · 209 ms · 111 buyers in 30 days

Money Mind

Money Mind describes static and rendered fetches with a second-render cross-check, confidence scoring and an attestation. The relevant text quote is $0.02. It says login, protected and CAPTCHA pages are refused without charge. Those are provider declarations about its process and billing, rather than an independently established accuracy measure. An attestation can identify an extraction record; it does not make the source page true or establish permission to reuse it.

1 web retrieval endpoint of 13 · $0.02/call (POST /v1/extract) · Base · uptime 30d 99.6% · 256 ms · 5 buyers in 30 days

ISHOB

ISHOB describes a robots-aware public-page fetch converted to Markdown, priced here at $0.001. Its header-only metadata quote is $0.0002. The table-extraction route accepts a public PDF or HTML page and is grouped with structured extraction at $0.001. That input distinction matters: a PDF-only converter would not qualify as a general page reader simply because a document has a URL.

3 web retrieval endpoints of 34 · $0.001/call (POST /v1/extract) · Base · uptime 30d 100% · 173 ms · 16 buyers in 30 days

Metalift

Metalift describes a scrape route with JavaScript rendering and escalation to proxies when required, plus URL mapping and sitemap routes. Its rendered-content reference is $0.002. The provider says billing follows the path used, so that catalog value does not establish the cost of every target page. Its mapping quote, $0.002, buys a URL inventory. Read the actual page before treating a discovered link as supporting evidence.

3 web retrieval endpoints of 4 · $0.002/call (POST /v1/scrape) · Base · uptime 30d 96.1% · 403 ms · 5 buyers in 30 days

agentutility

agentutility describes a static Cheerio scraper with extracted page text at $0.04. Its description does not establish a headless browser that executes a client application. Include a page whose useful content is already in the initial HTML when evaluating this route. A successful static fetch and a browser-rendered application are different acceptance cases, even when a provider uses the phrase server-side rendering.

1 web retrieval endpoint of 9 · $0.04/call (POST /scrape-website) · Base/SOL · uptime 30d 99.8% · 1502 ms · 36 buyers in 30 days

Zinin M2M Hub

Zinin M2M Hub lists URL-to-Markdown at $0.01 and structured extraction at $0.01. The latter describes deterministic extraction of fixed page fields such as metadata, JSON-LD, headings, links and detected contact or price strings. It should not be assumed to implement an arbitrary caller-defined semantic schema. Check the exact output keys against your application before replacing a configurable extractor with this route.

2 web retrieval endpoints of 12 · $0.01/call (POST /api/url-to-markdown) · Base · uptime 30d 100% · 178 ms · 3 buyers in 30 days

Reference cost and request budgets

The median selected reference is $0.005. At an unchanged request shape, 100,000 calls amount to $500.00.

Each service at the price of its main web retrieval endpoint, the one named in the table above.

Build a budget around accepted page reads. A browser retry, a second extraction pass or a separately billed crawl can change the total. A batch endpoint may price a bounded group rather than one URL, so record both requests and successful pages. Use the challenge for the precise target and mode when a provider describes escalation, upstream tolls or variable work.

Cache only according to the source and provider terms. Keep the extraction version with a cached record so a parser change can be distinguished from a source edit. A low metadata quote should not be multiplied as though it purchased rendered article text.

What the captured 402 establishes

The stored evidence for scrape402 documents the GET route with a public URL in the query. The illustrative target is example.com; no paid extraction was performed to produce this comparison.

curl -i 'https://x402.shizu.me/scrape?url=https%3A%2F%2Fexample.com'

HTTP/1.1 402 Payment Required

{
  "x402Version": 2,
  "resource": {
    "url": "https://x402.shizu.me/scrape"
  },
  "accepts": [
    {
      "asset": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
      "extra": {
        "name": "USD Coin",
        "version": "2"
      },
      "payTo": "0x217e5Fe265EB78b29067bF8324ef03a7D8e167C4",
      "amount": "6000",
      "scheme": "exact",
      "network": "eip155:8453",
      "maxTimeoutSeconds": 300
    }
  ]
}

402 of scrape402 for GET /scrape, as stored by the monitor (terms unchanged since 2026-09-27, last check 2026-10-26 17:43 UTC), with every accepted payment exactly as stored and without error, extensions, resource.tags, resource.mimeType, resource.description and resource.serviceName, which are not payment terms: the full body is also on the service page, under Payment evidence. Amounts are in the asset's smallest unit: USDC has 6 decimals, so 6000 is $0.006.

The generated reference price is $0.006, tied to the shown scrape path. A captured challenge establishes the advertised resource and payment requirements at that observation. It does not establish the quality of a paid response, prove that this target was extracted, or guarantee an unchanged future quote. Review the current challenge for your target URL before authorizing a payment.

How to read this comparison

The page starts with services matched to the web-scraping capability and filters their captured endpoints to the documented outputs described above. Inactive or degraded listings do not enter the operating comparison. Screenshot-only tools, raw HTML-string converters, SEO verdicts, general search routes and purchases of compute or API credits do not establish this retrieval task.

Price selection uses the same endpoint classification for rows, job candidates and inline references. The comparison date is October 26, 2026. Median monitoring uptime is 100%; monitoring response time belongs to the checks performed by the directory. It should not be interpreted as a benchmark of a long crawl or a successful browser extraction.

For repeatable evaluation, retain the request shape, endpoint, retrieval time, useful output and reason for rejection. Revisit the mapping when a provider changes its route contract. A renamed endpoint should be reviewed for the data it returns before it inherits an old type or price.

Questions

How to pay

Each service answers an unpaid call with HTTP 402 and its price; your agent signs a USDC payment on the listed network and retries. No API key, no account, no subscription. See what x402 is.