Eight providers fetch a page for you and hand back its contents: some unblock and render it, some run pre-built scrapers, some clean it up for a model.
Firecrawl to turn any website into clean Markdown for a model, from $16, and ScrapingBee or Bright Data to fetch the pages that fight back. Apify when a ready-made scraper for the site already sits in its marketplace.
A web scraping API takes a URL, or a target such as a product or a profile, and returns what is on the page. The work it saves is the work that breaks a home-made scraper: rotating proxies, rendering JavaScript, solving the challenge pages, retrying, and turning HTML into something a pipeline can use.
The directory carries seven web scraping APIs, and they sell three different things under one name. Bright Data and ScrapingBee unblock and render a page you name; Apify and Bright Data's Web Scraper API run pre-built scrapers that return structured records; Firecrawl, Tavily, Exa and Parallel fetch a page and clean it into text a model can read.
Which kind you need depends on whether you want the page, the record or the reading. Every price below is read from the vendor's own pricing page, and each provider's review carries the full rate card and its limits.
Most confusion about API and web scraping comes down to one thing. The difference between web scraping and API access is who agreed to the exchange. An API is a door the site built for you, with a documented format, a key and a rate limit. Scraping reads the page a person would see, and the site never promised it would stay the same shape.
So the question "API or web scraping" usually answers itself: use the site's own API when it has one that returns what you need, and scrape when it does not. A web scraping API sits between the two. It is an API to you, documented and keyed and billed, and scraping on the other side, against a site that offered no API.
That is why web scraping with API access to a provider is not the same as using the target's own API, and why the terms matter twice. The provider's terms govern your use of the provider, and the target site's terms govern the page. Neither is something this directory can rule on for you; read both before you scale a job.
Web scraping through API calls to a provider means the brittle parts become somebody else's problem, and the bill is how you pay for that.
The seven split cleanly once you ask what comes back. The same word covers a proxy network, a marketplace and a reader, and their prices are not comparable until you know which one you are buying.
A team scraping a few known sites wants the second kind, a team reading arbitrary company websites wants the third, and the first is for everything in between.
Most of what a web scraping API charges for is how hard the page fights. ScrapingBee publishes the ladder exactly: 1 credit for a rotating proxy with no JavaScript, 5 for the default rendered page, 10 or 25 for premium residential proxies, and 75 for stealth. Turning rendering off with render_js=false on a page that does not need it cuts a call from 5 credits to 1.
Bright Data is the proxy API for web scraping in its oldest form: a zone has a proxy username and password. A client that already speaks HTTP proxies can point at its superproxy and get the same unblocking as the API, billed at the same rate.
Firecrawl is the reminder that the ceiling is not always the price. Its own documentation says the real bottleneck is concurrent browsers, not requests a minute, because a rendered page holds a browser for seconds.
Before comparing rates, find out which configuration your target sites need, because the same provider can cost 75 times more per page on a hard site than on an easy one.
Three targets account for a lot of scraping demand, and each has a better answer than a generic fetch. An Amazon API for web scraping exists ready made: ScrapingBee publishes Amazon search and Amazon product APIs, and Bright Data sells an Amazon scraper among its pre-built ones.
A Google web scraping API is a SERP API by another name. Rather than rendering google.com yourself, use one of the providers on the SERP APIs page; ScrapingBee also sells its own Google results endpoint at 10 credits a call, and Apify routes Google results through its proxy at $2.50 to $1.70 per 1,000. A Google API for web scraping the open web, from Google itself, does not exist for a new customer.
A LinkedIn web scraping API is the most sensitive of the three. Bright Data's Web Scraper API returns LinkedIn records and bills per record delivered, so one call that returns 50 profiles is 50 units. LinkedIn's own terms are the ones to read first, and the LinkedIn API page covers what LinkedIn's own API allows and the providers that work from a profile URL instead.
For each of the three, a provider that already returns the record is cheaper and steadier than a scraper you point at the page and maintain yourself.
Most web scraping using API calls in Python is a plain HTTP request with a key, and every provider here works that way. Several publish a Python package as well, which is the shortest route to a python web scraping API that handles retries and pagination for you.
Apify's is apify-client, Bright Data's is brightdata-sdk, Parallel's is parallel-web and ScrapingBee's is scrapingbee, each installed with pip and each named on the vendor's own pages. Firecrawl, Tavily and Exa are called the same way through their REST APIs.
For web scraping with Python API clients at volume, the client matters less than the concurrency: ScrapingBee allows 25 requests in flight on its entry plan, Firecrawl 5 browsers, and Apify 32 concurrent runs. Size the pool to the plan, not to the machine.
Web scraping and API fundamentals in Python are the same skill here: send the request, read the status, back off on a 429.
Comparing web scraping API services on price means converting every unit into pages. These are the entry and volume rates each vendor publishes, the arithmetic done here.
The cheapest web scraping service API on paper is the one whose cheapest configuration your targets actually allow, so test your own sites before you choose a plan.
One of the seven is open source. Firecrawl's code is public, so the behavior the hosted service charges for is inspectable, and self-hosting it is a real option when the constraint is data residency rather than engineering time.
The hosted plan still earns its keep for most teams, because the proxies, the browser fleet and the unblocking are the expensive part, and a self-hosted copy brings none of them. A web scraping API open source build is a starting point, not the service.
If self-hosting is the plan, price the proxies you will need before comparing it with a $19 hosted plan.
Entry price and unit cost are read from each vendor's own pricing page. The unit differs from row to row, a credit, a request, a compute unit or a search, so convert to pages before comparing.
| Provider | Returns | Entry price | Unit cost | MCP server |
|---|---|---|---|---|
| Exa | Firmographics, Web pages | $4/1k searches | $0.001 per page, Contents | Yes |
| Firecrawl | Firmographics, Technographics, Web pages | $16/mo annual | $0.0032 per credit | Yes |
| Tavily | Firmographics, Web pages | $0.008/credit | $0.008 per credit | Yes |
| Apify | Firmographics, Web pages | $19/mo | $0.20 per compute unit | Yes |
| Bright Data | Firmographics, Job postings, LinkedIn profile, Web pages | $1.50/1K requests | $0.0015 per request | Yes |
| ScrapingBee | Firmographics, Web pages | $19/mo | $0.00025 per credit | Yes |
| Perplexity | Web pages | $5/1k searches | $0.005 per search request | Yes |
| Parallel | Firmographics, Web pages | $1/1k searches | $0.001 per page, Extract | Yes |
Read from each vendor's own pages on 24 October 2026, and each review carries the full rate card.
The 8 providers are chosen by the data each one returns, read off its record, not by hand. Billing model is the filter worth starting with, because it decides whether a quiet month costs you anything.
Each has a full review, with the rate card read off the vendor's own page: Exa, Firecrawl, Tavily and Apify.
Also with a full review: Bright Data, ScrapingBee, Perplexity and Parallel.
Exa is what you reach for when the signal you need is not in anybody's database yet. It returns clean page content rather than links, which is the difference between an agent that can answer a research question and one that hands you ten tabs.
Firecrawl takes a URL and gives you back the page as Markdown, with the navigation, the cookie banner and the JavaScript already dealt with. It is the layer most GTM agents are missing: not a database of companies, but the ability to read any company's own site.
Tavily is a search API built for language models rather than for people, and priced simply enough that you can model it on a napkin: one credit a basic search, two for a deep one, $0.008 a credit. The free tier is 1,000 credits a month with no card. Tavily now trades as part of Nebius.
Apify runs other people's scrapers. Thousands of Actors in Apify Store, from Google Maps to whole-site crawlers, are started, polled and read through one REST API and one token. The plan you buy is prepaid usage in dollars, and what a single scrape costs is set by whoever wrote the Actor.
Bright Data is web data infrastructure: proxies, an unblocker, a SERP API and more than 1,500 pre-built scrapers. The three API products a GTM team calls, the SERP API, the Web Unlocker API and the Web Scraper API, are priced the same way: $1.50 per 1,000 successful requests or records, or $499 a month for about 380,000.
ScrapingBee is one scraping endpoint that renders JavaScript and rotates proxies for you, priced in credits whose cost depends on the configuration: 1 for a plain fetch, 5 for the default rendered page, up to 75 for stealth. The plans are generous on credits, and the ceiling that actually binds is concurrency.
The Perplexity API is now two products: a Search API that returns ranked results and extracted text at $5 per 1,000 requests with no token costs, and an Agent API that runs third-party models with web search as a priced tool. Sonar Chat Completions support ended on September 27, 2026.
Parallel prices everything per request rather than per token, so a query's cost is known before it runs. Search is $1 or $5 per 1,000 depending on the mode, and the Task API for structured research runs from $5 to $2,400 per 1,000. The default search mode is the $5 one.
In the order they cost you money.
Decide whether you want the raw page, a structured record from a known site, or clean text for a model. That choice picks the kind of provider before price does.
A plain fetch and a stealth render can differ by 75 times on one provider. Test your own sites, because the configuration they need sets the real rate.
Rendered pages hold a browser for seconds, so the number of requests in flight caps throughput long before any per-minute limit. Compare plans on concurrent requests or browsers.
ScrapingBee and Bright Data charge for successful calls only, and ScrapingBee does not bill a 429. A provider that bills a blocked attempt turns a hard site into a bill.
The same providers, seen three other ways: head to head on the compare hub, by the reason teams leave on the alternatives hub, and wired together with the cost run end to end on the stacks hub.
The same providers, sorted by other data: SERP APIs.
By the stage of the motion they serve: Signal APIs and Sourcing APIs.
Head to head, with the arithmetic done: Exa vs Tavily, Apify vs Firecrawl, Apify vs Bright Data and Apify vs ScrapingBee.
Also side by side: Firecrawl vs ScrapingBee, Exa vs Perplexity and Exa vs Parallel.
Teams leaving one of these have somewhere to go: Firecrawl, Exa and Tavily.
And: Apify, Bright Data and Perplexity.
Clay runs a waterfall across 150+ data providers from one table, without a separate contract with each. A lookup that finds nothing costs no data credit, and the free plan includes 500 actions a month.
Frontal is a Clay Elite Studio Partner.
8, chosen by what each one returns rather than by hand, and each with its own review carrying the price, the rate limit and the auth method read from the vendor's own pages.
Exa, Firecrawl, Tavily, Apify, Bright Data, ScrapingBee and Parallel. What each free tier includes differs a lot, so the review says what you actually get.
Of the 3 that price per credit, ScrapingBee at $0.00025. The others meter a different unit, so their rates are not comparable number for number, and a monthly floor or a charge on a miss changes the bill either way.
Every one of them meters you, so the unit cost is the comparable figure once the unit is the same, and the monthly floor is what you actually pay at low volume.
Exa, Firecrawl, Tavily, Apify, Bright Data, ScrapingBee, Perplexity and Parallel. The ones the directory reviews live on the MCP page with the connect details.
An API is an interface the site built and documents for you; scraping reads the page a person would see. A web scraping API gives you an API on your side and does the scraping on the other, against sites that offer no API.
That is most of what it is for. Bright Data's Web Unlocker handles CAPTCHAs, fingerprinting and retries, and ScrapingBee renders JavaScript by default, which is why its default call costs 5 credits rather than 1.
Firecrawl, which turns any URL into Markdown with the navigation and cookie banner removed. Tavily and Parallel extract page content, and Exa's Contents endpoint returns page text at $1 per 1,000 pages.
This directory does not give legal advice. The provider's terms cover your use of the provider and the target site's terms cover the page, so read both, and the target site's first.
