Web scraping API pricing

What a web scraping API is

A web scraping API accepts a URL and returns the received page, rendered page, or extracted response without requiring the buyer to operate the proxy and browser layer directly. The defining job is retrieval from a URL. Some products add rendering, residential routing, screenshots, or structured extraction, but those are modifiers to the retrieval job rather than proof that every row returns the same output. ScraperAPI documents URL-based requests and separate credit costs for rendering and target-specific modes, while Zyte API separates HTTP and browser-rendered retrieval and optional extraction. (ScraperAPI credit documentation, checked Jul 26, 2026; Zyte API pricing, checked Jul 26, 2026.)

A cloud browser is the neighbouring category most often confused with a scraping API. A scraping API is a request-and-response abstraction: send the target and options, then receive a result. A cloud browser exposes a browser session that the buyer’s code or agent controls over time. Browserbase publishes browser-hour, proxy, Search, Fetch, and Extract meters, while ScraperAPI publishes request-credit rules. That difference changes both architecture and budgeting. Choose a cloud browser when the workflow must preserve session state or execute a sequence of interactions. Choose a scraping API when the primary unit is an independently retrievable URL. (Browserbase pricing, checked Jul 25, 2026; ScraperAPI pricing, checked Jul 26, 2026.)

A SERP API is narrower than a general scraping API because its contract begins with a search query and returns search-engine data. A web search API also begins with a query, but is usually designed for discovery or grounding rather than faithful capture of a named search-engine results page. AI extraction is downstream transformation: it converts a received page or browser state into a requested schema, answer, or agent-ready representation. Products can span these categories, which is why the same product can legitimately appear on more than one comparison page. (SerpApi pricing, checked Jul 26, 2026; Brave Search API, checked Jul 26, 2026; Firecrawl extraction guide, checked Jul 26, 2026.)

What this category bills on

The most direct meter is a request or a successful result. Credit-based products translate the request into credits, then apply different credit weights for JavaScript rendering, premium proxies, protected targets, screenshots, or other modes. Result-based products may bill only when they return a qualifying result, while request-based products can define chargeable status codes or cancellation rules differently. The buyer therefore needs both the plan allowance and the exact multiplier for the intended target and options. (ScraperAPI credit documentation, checked Jul 26, 2026; Oxylabs pricing, checked Oct 9, 2026; ScrapingBee FAQ, checked Jul 26, 2026.)

Not every product in this category exposes a page-request billing unit. Apify meters platform compute, storage, transfer, proxy use, and Actor-specific events. Cloudflare Browser Rendering and Lightpanda expose browser-time meters. Spider.cloud combines compute or concurrency with bandwidth. Those products remain part of the scraping market because they can perform the retrieval job, but their public meters cannot be converted into the table’s page workload without inventing session duration, memory, bandwidth, or concurrency assumptions. (Apify pricing, checked Jul 26, 2026; Cloudflare Browser Rendering pricing, checked Jul 25, 2026; Lightpanda pricing, checked Jul 25, 2026; Spider.cloud pricing, checked Jul 26, 2026.)

The comparison workload

The table uses one JavaScript-rendered page request as the unit and shows every product carrying scraping-api in the dataset. It does not silently add a premium proxy, residential route, target-domain surcharge, screenshot, AI extraction step, browser duration, or guessed retry rate. An assumption marker means a required modifier is not explicit in the product record and is provisionally treated as included. A not-computable or quote-required row is retained because an unpublished or incompatible meter is market information, not a reason to omit the product. (Scraping API Prices methodology, checked Jul 26, 2026.)

Rows are ordered by evidence state before bill size: assumption-free calculations, assumption-dependent calculations, published lower bounds, quote-required rows, and rows whose public meter does not fit the workload. Affiliate data does not affect inclusion or order. An assumption-marked row is a question to resolve with the vendor, not evidence for a cheapest claim. (Scraping API Prices methodology, checked Jul 26, 2026.)

What can change the bill

Target, rendering mode, proxy route, and output type can change the number of credits or the applicable meter. Pagination and detail-page enrichment create additional retrievals. Media transfer, storage, browser time, compute, and extraction tokens can be separate charges on hybrid products. Concurrency is usually a capacity limit rather than a usage billing unit, so a plan can fit the request count while failing the required completion window. (ScraperAPI credit documentation, checked Jul 26, 2026; Apify pricing, checked Jul 26, 2026; Browserbase pricing, checked Jul 25, 2026.)

How these charges fit the workload

• Oxylabs Web Scraper API publishes a JavaScript custom-domain rate that maps to the kept pages in this workload ($1.50 per 1,000 on its $30 Starter, $1.40 on Basic, $1.30 on Flexible). Its subscriptions now cover every Oxylabs product, and each is a balance drawn down at those rates that can be topped up by any amount during the month at the plan’s own rate. The cheapest way to buy the volume is $30 Starter at 10,000 JavaScript pages, $100 Basic plus a $40 top-up ($140) at 100,000, and $500 Flexible plus an $800 top-up ($1,300) at 1,000,000. Checked Oct 9, 2026; top-up rule from Oxylabs top-ups, checked Oct 13, 2026.
• Crawlbase bills successful requests from a graduated pay-as-you-go card or from monthly credit blocks. JavaScript rendering doubles a request, and each website’s difficulty tier multiplies it from ×1 to ×20, so the table gives its bill as a range across the eight tiers, from the easiest sites to the hardest. Checked Oct 11, 2026.
• Zyte API puts each website in one of five difficulty tiers, with separate prices for HTTP and browser-rendered requests and optional monthly commitments, so the table gives its bill as a range across the five tiers. Failed responses are free. Checked Oct 11, 2026.
• Apify runs scraping jobs as Actors and meters platform resources rather than publishing a generic rendered-page billing unit. The pricing page is the route to the entry tier or a custom platform contract. Checked Jul 26, 2026.
• Cloudflare Browser Rendering exposes browser automation through Workers and bills browser usage rather than a generic page request. Its public documentation is the purchase path; no separate enterprise browser offer is published there. Checked Jul 25, 2026.
• Diffbot publishes a credit mapping for its Extract operation. The table can now price that operation when it matches the selected workload; crawling, natural-language processing and Knowledge Graph operations retain their separate meters. Checked Oct 5, 2026.
• Lightpanda Cloud provides hosted browser execution with browser-time allowances and a bring-your-own-proxy model. Its pricing page is the route to the entry tier or a custom deployment and volume agreement. Checked Jul 25, 2026.
• ScrapeGraphAI combines scraping formats, structured extraction, search, crawling, monitoring, and optional stealth under shared credits. Its product page is the route to the entry tier or custom credits and limits. Checked Jul 26, 2026.
• SOAX Web Data API offers request allowances with rendering, proxy routing, and anti-bot handling presented as included, but the stored request meter has no billing unit suitable for this calculation. Its pricing page routes buyers to custom request volume and terms. Checked Jul 26, 2026.
• Spider.cloud provides crawling and browser-capable retrieval through bandwidth-and-compute or concurrency-based offers. Its pricing page is the route to a public plan or custom arrangement, but neither model converts to a page bill without workload assumptions. Checked Jul 26, 2026.

How to use the comparison

Start with the row whose meter matches the intended job, then verify the target, rendering, routing, output, failure-billing rule, concurrency, and contract term in writing. Recalculate whenever one of those inputs changes. Treat not-computable as a procurement question and quote-required as a contract path. Neither state means free, unavailable, or outside the market. (Scraping API Prices methodology, checked Jul 26, 2026; ScraperAPI credit documentation, checked Jul 26, 2026.)