# ScrapingBee

> ScrapingBee is a web scraping API. You send it a URL and it returns the page content, handling the parts that normally break a scraper: rotating proxies, headless Chrome rendering, anti-bot evasion, geotargeting, and retries. It also offers structured extraction through CSS or XPath rules and an AI query mode, plus dedicated endpoints for Google Search, Amazon, and Walmart. Pricing is credit-based, and credits are only consumed on successful responses.

- Category: GTM Engineering (https://saastracker.org/categories/gtm-engineering)
- Website: https://www.scrapingbee.com
- Starting price: $19.99 per month (Hobby, 75,000 credits, 25 concurrent requests)
- Free plan: No
- Free trial: 1,000 free API credits, no credit card required
- Founded: 2019, HQ: France, Ownership: Acquired by Oxylabs (June 2025)
- Profile last reviewed: 2026-08-23
- Canonical profile: https://saastracker.org/products/scrapingbee

## Overview

Every team that has tried to build data collection in house arrives at the same wall. The first version, a Python script with requests and BeautifulSoup, works for a week. Then the target site starts serving content through JavaScript, or begins returning 403 to datacenter IPs, or fingerprints the TLS handshake, and the script becomes a permanent maintenance job for someone who was hired to do something else. ScrapingBee exists to make that wall someone else's problem: one HTTP endpoint, one API key, and the proxy pool, browser farm, and evasion logic sit behind it.

The design choice that defines the product is that it stays an API rather than becoming a platform. There is no visual scraper builder, no scheduler, no hosted storage of results, and no dataset marketplace. You still write the crawl loop, the parsing, the deduplication, and the storage; ScrapingBee handles fetching. That narrowness is why developers like it and why non-technical buyers usually should not buy it. The surface area is small enough to learn in an afternoon and stable enough that integrations rarely break.

Around that core the company has added the pieces that make the API usable for modern work: a JavaScript scenario language for clicking, scrolling, and filling forms before extraction; a markdown mode that returns LLM-ready text instead of raw HTML; an ai_query parameter that describes wanted fields in plain language rather than selectors; and an official MCP server so an agent can fetch live pages directly. For go-to-market engineering specifically, this is the layer that feeds a Clay table, an enrichment pipeline, or a competitive-pricing monitor with data no vendor sells.

The commercial context changed in June 2025, when Oxylabs acquired ScrapingBee in an eight-figure all-cash deal. ScrapingBee had been mostly bootstrapped (TinySeed was the only outside money), had reached roughly $5 million ARR with a team of about six, and continues to operate as a distinct product rather than being folded into the Oxylabs enterprise line. The practical effect so far has been infrastructure improvements and credit-cost reductions on some endpoints, not a repricing of the self-serve plans.

## How it works

1. You make a GET request to the API endpoint with your key and a URL-encoded target. By default the request runs through a headless Chrome instance with JavaScript rendering enabled and returns the resulting HTML. Official clients exist for Python, Node, PHP, Ruby, Go, and Java, but nothing about the API requires an SDK; curl works fine, which is why it drops into any language or automation tool without ceremony.

2. Parameters control the expensive parts. render_js toggles the browser (set it false and the request is a plain fetch at one credit). premium_proxy routes through residential IPs for sites that block datacenter ranges. stealth_proxy is the last resort for aggressively defended targets. country_code pins geography, which matters for pricing and search results that vary by market. wait, wait_for, and wait_browser control when the page is considered ready.

3. For interactive pages, js_scenario accepts a JSON array of browser instructions: click a selector, scroll, fill a field, wait, evaluate JavaScript. This is how you get past a cookie wall, expand a lazy-loaded list, or step through a paginated table before the HTML is captured, without running your own Playwright fleet.

4. Extraction happens in three ways. extract_rules maps CSS or XPath selectors to JSON keys, so the API returns structured data instead of a document you still have to parse. ai_query and ai_extract_rules describe the wanted fields in natural language and let a model do the selector work, which survives layout changes that would break a hand-written selector. A markdown or plain-text mode returns clean prose for anything headed into an LLM context window.

5. Billing runs on credits, deducted per request according to what you turned on: one credit for a plain fetch, five for JavaScript rendering, ten for a premium proxy without rendering, twenty-five for premium proxy plus rendering, and seventy-five for stealth proxy. Only successful responses (HTTP 200, 404, or 410) are charged; a timeout or a 500 costs nothing, which is a genuinely material difference from vendors that bill every attempt.

## Best for

Developers and go-to-market engineers who need reliable page fetching as a component inside their own pipeline: a growth team enriching accounts, a pricing analyst monitoring competitors, an agency building a lead-sourcing script, or an AI product that needs fresh web content on demand.

## Not the right fit for

- Non-technical users who want to point at a page and click through a wizard; there is no visual builder, and the whole product assumes you write code.
- Teams that want the scraper to run on a schedule and store results; ScrapingBee fetches, it does not orchestrate, host, or persist anything.
- Anyone scraping sites that require a logged-in account with sticky, long-lived sessions, where a session-management platform or a purpose-built automation tool fits better.
- Buyers who want proxies by the gigabyte rather than by the request; the credit model is per-call, and heavy bandwidth workloads are cheaper on a raw proxy network.
- Very spiky or seasonal workloads, since credits reset monthly and do not roll over, so a quiet month is money spent on nothing.

## Features

### Fetching and rendering

The core job: get the page, the way a real browser would see it.

- **HTML API**: A single GET request returns the page source after rendering, with retries, redirects, and header management handled server-side.
- **Headless Chrome rendering**: JavaScript-built pages, single-page apps, and lazy-loaded content resolve before capture, which is the failure mode that kills most homegrown scrapers first.
- **JavaScript scenarios**: A JSON instruction list for clicking, scrolling, typing, waiting, and evaluating script in the page before the HTML is returned, covering cookie walls and paginated tables.
- **Wait controls**: Fixed delays, wait-for-selector, and browser lifecycle events (domcontentloaded, load, networkidle) let you decide precisely when a page counts as ready.
- **Resource blocking**: Block ads, images, fonts, or arbitrary resource types to cut render time and avoid paying latency for assets you will never parse.
- **Screenshots**: Full-page, viewport, or single-element screenshots of any URL, useful for visual regression checks and for evidencing competitor pricing changes.
- **Device and viewport emulation**: Request the mobile or desktop rendering of a page, which matters when the two layouts serve materially different content or prices.

### Proxies and evasion

The layer that decides whether a request comes back at all.

- **Rotating proxy pool**: IP rotation is automatic and included in the credit cost, with no separate proxy contract, bandwidth meter, or pool to configure.
- **Premium residential proxies**: A residential tier for sites that reject datacenter ranges outright, enabled with one boolean at a higher credit cost per request.
- **Stealth proxy mode**: The most aggressive configuration for heavily defended targets, priced at seventy-five credits so that it is a deliberate choice rather than a default.
- **Auto-mode**: Lets the API escalate through configurations until one succeeds, charging for the level that worked and nothing at all if every attempt fails.
- **Geotargeting**: Country-level targeting via an ISO country code, which is how you see local pricing, local search results, and geo-gated catalogs.
- **Session persistence**: A session_id parameter keeps the same IP across a sequence of requests, which is what multi-step flows need to avoid tripping consistency checks.
- **Custom headers and cookies**: Forward your own headers and cookies through the proxy, so authenticated or preference-dependent pages behave as they would in your browser.

### Extraction and AI output

Getting structured data back instead of a document to parse yourself.

- **CSS and XPath extract rules**: A JSON map from selectors to output keys, returned as clean JSON, so simple jobs need no parsing library on your side at all.
- **AI data extraction**: Describe the fields you want in plain language with ai_query or ai_extract_rules and let a model locate them, which tolerates layout changes that would break a fixed selector.
- **Markdown and plain-text mode**: Returns readable text stripped of navigation and boilerplate, sized for an LLM context window rather than for a parser.
- **Nested and list extraction**: Extract rules support repeated blocks and nested objects, so a product grid comes back as an array of records rather than one flat blob.
- **JSON response mode**: Response metadata including status, resolved URL, cost in credits, and headers is available alongside the payload for logging and cost attribution.

### Dedicated endpoints and integrations

Pre-built targets and the surfaces that let non-code tools call the API.

- **Google Search API**: A structured SERP endpoint returning organic results, ads, and related queries with locale control, priced per call rather than requiring you to parse Google yourself.
- **Retail endpoints**: Purpose-built Amazon and Walmart endpoints return normalized product, price, and review fields, including store and postcode localization for Walmart.
- **AI answer endpoints**: Endpoints that capture ChatGPT and Gemini answers for a query, which is how teams now track whether their brand appears in model-generated recommendations.
- **MCP server**: An official Model Context Protocol server so Claude Code, Cursor, and other agents can fetch, screenshot, and extract from live pages without custom glue code.
- **No-code connectors**: Native steps in Make, n8n, and Zapier let the API be called from an automation without writing a request by hand.
- **Official SDKs and CLI**: Maintained clients for Python, Node, PHP, Ruby, Go, and Java, plus a command-line tool for one-off fetches and debugging.

## Use cases

- **Go-to-market engineer building an enrichment pipeline**: An outbound team wants to score accounts on signals no data vendor sells: hiring pages, pricing page changes, and which integrations a prospect lists on its site. Outcome: A nightly job calls ScrapingBee with extract rules per page type, writes JSON into the warehouse, and syncs into Clay and the CRM, giving reps a signal column that no competitor buying the same enrichment vendor has.
- **Ecommerce operator monitoring competitor pricing**: Three rival storefronts change prices unpredictably, block datacenter IPs, and render prices client-side, so the intern with a spreadsheet has stopped being viable. Outcome: A twice-daily script hits each product URL with premium proxies and geotargeting, stores price history, and alerts on movement, at a credit cost that is a small fraction of a repricing platform subscription.
- **Founder of an AI product that needs live web content**: The application must read arbitrary user-supplied URLs and hand clean text to a model, but a third of real-world pages fail with a plain HTTP fetch. Outcome: The fetch layer becomes one API call in markdown mode, success rates rise sharply on JavaScript-heavy sites, and no engineer is asked to maintain a browser fleet before product-market fit.
- **SEO agency tracking rankings across client markets**: Clients want position tracking in several countries, and the agency does not want to pay a per-keyword rank tracker for exploratory queries. Outcome: The Google Search endpoint returns structured SERPs per locale on demand, feeding a custom dashboard where the marginal cost of another query is credits rather than a plan upgrade.

## Pricing

Credit-based subscription. Every plan is a monthly credit allowance plus a concurrency limit, and each request deducts credits according to the options enabled: one credit for a plain fetch, five with JavaScript rendering, ten for a premium proxy alone, twenty-five for premium proxy plus rendering, and seventy-five for stealth proxy. All features are available on every paid tier; higher tiers buy volume, concurrency, and support, not capability.

- **Free trial**: $0 one time. 1,000 API credits with no card required; Every feature unlocked, including premium and stealth proxies; Enough for roughly 200 rendered requests, or 1,000 plain fetches. A trial, not a free tier; once the credits are gone the account needs a plan.
- **Hobby**: $19.99 per month. 75,000 credits (about 15,000 rendered requests); 25 concurrent requests; JavaScript rendering, rotating proxies, geotargeting, screenshots.
- **Freelance**: $49.99 per month. 250,000 credits (about 50,000 rendered requests); 50 concurrent requests; The usual starting point for a real production job rather than an experiment.
- **Startup**: $99.99 per month. 1,000,000 credits; 100 concurrent requests; The point where the effective price per rendered request drops meaningfully.
- **Business**: $249.99 per month. 3,000,000 credits; 200 concurrent requests; Priority email support, a dedicated account manager, and team management. Business+ sits above at $599.99 for 8,000,000 credits and 400 concurrent requests.
- **Enterprise**: From $999.99 per month. Published enterprise steps from 14,000,000 credits up to 120,000,000 credits; 500 to 900 concurrent requests depending on the step; Custom contracts above the largest published step. Unusually, the enterprise tiers are published with prices rather than hidden behind a call.

Billing notes:

- Credits are only deducted for successful responses, defined as HTTP 200, 404, or 410; timeouts and server errors cost nothing, which matters a great deal when scraping defended targets.
- Credits do not roll over. Unused allowance expires at the end of the billing month, so a seasonal workload wastes a real share of what it pays for.
- The credit multipliers, not the headline allowance, determine actual capacity. A 250,000-credit plan is 250,000 plain fetches, 50,000 rendered requests, or about 3,300 stealth-proxy requests.
- Pricing is published monthly on every tier including enterprise, with no annual-commitment discount advertised as of August 2026.
- Plans can be cancelled from the dashboard without contacting sales, and running out of credits mid-cycle is handled by upgrading or renewing early rather than by an automatic overage charge.
- Concurrency, not just credits, caps throughput. A large one-off crawl on a low tier finishes slowly even with credits to spare.

Value assessment: Judged against the alternative of running your own proxy contract plus a Playwright fleet, ScrapingBee is cheap: the Freelance plan costs less per month than a couple of hours of the engineer who would otherwise be maintaining browser infrastructure. Judged against raw bandwidth-priced proxy networks, it is expensive, because you are paying per request for rendering you may not need. The honest rule is that ScrapingBee wins when a meaningful share of your targets need a real browser or residential IPs, and loses when they do not. The successful-requests-only billing and the published enterprise pricing both improve the deal relative to competitors who bill every attempt and hide the top of the range.

## Strengths

- One endpoint replaces the three hardest pieces of a scraping stack: proxy rotation, headless browsers, and anti-bot handling.
- Billing only on successful responses, which removes the worst cost surprise in the category.
- Documentation and error messages are unusually clear, and the API is small enough to learn in an afternoon.
- Every feature is available on the cheapest paid plan; higher tiers buy volume and support rather than unlocking capability.
- Published prices all the way up the enterprise range, which is rare in web data and makes budgeting possible without a sales call.
- AI extraction and markdown output make it usable as the fetch layer for an LLM application without a separate cleaning step.
- An official MCP server and native Make, n8n, and Zapier steps put the API inside agent and automation workflows without custom glue.

## Limitations

- The credit multipliers make effective cost hard to forecast. A workload that quietly shifts from rendered to stealth requests multiplies its bill by fifteen with no change to the code except one boolean.
- Credits expire monthly with no rollover, which penalizes bursty and seasonal scraping directly.
- It is a fetch API only: no scheduler, no hosted storage, no dataset browser, no visual builder, so a non-developer cannot use it at all.
- Concurrency limits are tied to plan tier, so a large backfill on a low plan is throughput-bound even when credits are available.
- Sticky sessions exist but long-lived authenticated crawling is not the product's strength; account-based scraping usually needs a different tool.
- For bandwidth-heavy workloads that do not need rendering, per-request credit pricing is worse value than a gigabyte-priced proxy network, including the one now owned by its parent company.
- The team was around six people at acquisition, so support depth and roadmap velocity depend on how Oxylabs resources it going forward.

## Comparisons

- **ScrapingBee vs Zenrows**: The most direct comparison in the category: both are request-priced scraping APIs with JavaScript rendering, residential proxies, and anti-bot bypass. ZenRows leans harder on advertised bypass capability and bundles a scraping browser and proxy products alongside the API. ScrapingBee is the more conservative, better-documented option, with successful-requests-only billing and published prices to the top of the range. Teams whose targets are the hardest-defended sites tend to test ZenRows; teams who want predictable behavior and clear docs tend to stay on ScrapingBee.
- **ScrapingBee vs Octoparse**: Different buyers entirely. Octoparse is a desktop and cloud point-and-click scraper: you record a workflow in a visual builder, schedule it, and download results, with no code required. ScrapingBee is an API with no interface at all. If nobody on the team writes code, Octoparse is the only one of the two that is usable; if you already have a pipeline and only need the fetch step to stop failing, Octoparse's builder is overhead you would route around.
- **ScrapingBee vs Firecrawl**: Firecrawl is aimed squarely at LLM ingestion: crawl a whole site, get clean markdown, with crawl orchestration built in. ScrapingBee is a per-URL fetcher whose markdown mode covers the same ground for single pages but leaves the crawl loop to you. Firecrawl is the better default when the job is 'ingest this documentation site into a knowledge base'; ScrapingBee is better when the job is targeted extraction from defended pages where proxy quality decides success.
- **ScrapingBee vs Apify**: Apify is a platform rather than an API: a marketplace of prebuilt actors, hosted execution, scheduling, storage, and a proxy layer, priced on compute and usage. ScrapingBee does one slice of that and does not host anything. Choose Apify when you want a ready-made scraper for a known target and somewhere to run it; choose ScrapingBee when you have your own infrastructure and want to keep it, paying only for the fetch.
- **ScrapingBee vs Serper**: Serper is a fast, very cheap Google SERP API and nothing else. If search results are the only thing you need, Serper is the lower-cost specialist and ScrapingBee's Google endpoint is not worth the credit spend. ScrapingBee earns its place when SERPs are one input among many and the same key also needs to fetch arbitrary product pages, retail listings, and JavaScript-heavy sites.
- **ScrapingBee vs PhantomBuster**: PhantomBuster automates actions inside logged-in accounts, principally social platforms, with scheduled prebuilt phantoms and no code. ScrapingBee fetches public pages through an API and deliberately does not do session-based platform automation. Growth teams frequently run both: PhantomBuster for the account-bound social work, ScrapingBee for the open web that feeds the same enrichment table.

## Implementation

- Setup time: Minutes. Sign up, copy the API key, and change the URL in an existing HTTP call. Getting a difficult target working reliably takes longer, usually an hour or two of tuning render options, wait conditions, and proxy level against real responses.
- Learning curve: Low for a developer and impassable for anyone else. The parameter set is small and well documented; the real skill is cost discipline, knowing when a page genuinely needs rendering or a residential IP rather than defaulting to the expensive configuration everywhere.
- Onboarding: Entirely self-serve. Documentation, SDKs in six languages, a large tutorial library covering scraping in specific stacks, and a knowledge base. Dedicated account management appears only at the Business tier and above.
- Migration: Moving from another scraping API is mostly a change of base URL and parameter names, since the request shapes are similar across vendors. The real migration work is recalibrating cost, because credit multipliers and the definition of a billable request differ enough between providers that a like-for-like volume estimate will be wrong. Run a week in parallel on a sample of your hardest targets and compare cost per usable result, not cost per request.

## Platform, API & security

- Platforms: REST API over HTTPS, Official SDKs: Python, Node.js, PHP, Ruby, Go, Java, Command-line client, MCP server for AI agents, Make, n8n, and Zapier connectors
- API: The API is the product. A single GET endpoint takes an API key and a URL plus optional parameters for rendering, proxy level, geography, waiting, extraction rules, AI queries, screenshots, and sessions. Separate endpoints cover Google Search, Amazon, Walmart, and AI answer capture. Responses return the payload plus metadata including the credit cost of the request.
- Compliance: GDPR
- Data residency: Operated from France under EU data protection law; the parent company, Oxylabs, is based in Lithuania. ScrapingBee does not store scraped page content as a product feature, so residency questions center on request logs rather than results.
- SSO: Team management is available from the Business tier; single sign-on is handled through enterprise arrangements rather than published as a self-serve feature.
- Security notes: API keys are the sole credential and should be held server-side, since a leaked key is directly spendable in credits. Requests are made from ScrapingBee infrastructure, so target sites see the vendor's proxy IPs rather than yours, which is a benefit for evasion and a consideration for any target whose terms you are bound by. Scraping legality depends on jurisdiction, the data, and the site's terms; the tool does not make that determination for you.

## Support

- Channels: Email support on all plans, Priority email support from the Business tier, Dedicated account manager from the Business tier, Knowledge base and status page
- Documentation: Thorough and practical: a complete parameter reference, worked examples per SDK, and a large tutorial library covering scraping patterns in Python, Node, Go, PHP, and others. The tutorials rank well enough that they function as the company's main acquisition channel.
- Community: No formal user forum. The public presence is the blog and tutorial archive plus the founders' long history of writing openly about the business, which has made ScrapingBee a widely cited reference point in developer discussions of scraping infrastructure.

## Company

- Founded: 2019
- Founders: Kevin Sahin, Pierre de Wulf
- Headquarters: France
- Ownership: Acquired by Oxylabs (June 2025)
- Employees: ~6 at acquisition (2025)
- Funding: Largely bootstrapped; the only outside capital was a TinySeed investment in 2020. Reached roughly $5 million ARR before an eight-figure all-cash acquisition.

Timeline:

- 2019: Kevin Sahin and Pierre de Wulf launch ScrapingBee in France after two earlier startups failed, selling a scraping API rather than a scraping service.
- 2020: Joins the TinySeed accelerator, the company's only outside funding.
- 2021: Passes $1 million in annual recurring revenue, still with a very small team.
- 2023: Adds AI-assisted extraction so fields can be described in natural language instead of CSS selectors.
- 2025: Oxylabs acquires ScrapingBee in an eight-figure all-cash deal at roughly $5 million ARR and 2,500 customers; the product continues to operate separately.
- 2026: Ships an official MCP server for AI agents, adds AI answer-capture endpoints, and reduces the credit cost of the Google Search endpoint.

## Integrations

Make, n8n, Zapier, Claude Code, Cursor, Python, Node.js, Go, PHP, Ruby, Java, Google Sheets (via automation platforms)

## FAQ

### What is ScrapingBee used for?

Fetching web pages that resist ordinary HTTP requests. You send a URL to its API and get back the rendered HTML, clean markdown, or structured JSON, with rotating proxies, headless Chrome, geotargeting, and anti-bot handling done on the vendor's side. Typical jobs are competitor price monitoring, lead and account enrichment, SERP collection, and feeding live web content to an AI application.

### How much does ScrapingBee cost?

Paid plans start at $19.99 per month for the Hobby tier (75,000 credits, 25 concurrent requests), then $49.99 for Freelance, $99.99 for Startup, $249.99 for Business, and $599.99 for Business+. Enterprise steps are published from $999.99 up to $5,799.99. New accounts get 1,000 free credits with no card required, but there is no permanent free plan.

### How does the ScrapingBee credit system work?

Each request costs credits based on what you enable: one credit for a plain fetch with rendering off, five for a JavaScript-rendered request, ten for a premium residential proxy without rendering, twenty-five for premium proxy plus rendering, and seventy-five for stealth proxy mode. This is the number that matters, not the headline allowance. Credits are only deducted for successful responses, meaning HTTP 200, 404, or 410.

### Do ScrapingBee credits roll over?

No. Unused credits expire at the end of the billing month. If you run out mid-cycle you can upgrade or renew the subscription early rather than being charged an automatic overage. For workloads that spike a few times a year this is a real cost, since the plan has to be sized for the peak month.

### Does ScrapingBee charge for failed requests?

No. Only responses with HTTP status 200, 404, or 410 consume credits. Timeouts and server errors are free, and in auto-mode, where the API escalates through proxy configurations until one works, a request that fails at every level costs nothing. This is one of the clearer differences from providers that bill every attempt.

### Is ScrapingBee still independent, or was it acquired?

It was acquired by Oxylabs in June 2025 in an eight-figure all-cash deal, having been mostly bootstrapped to about $5 million ARR with a team of roughly six. It continues to operate as a separate product rather than being absorbed into the Oxylabs enterprise line, and self-serve pricing has not been restructured since the deal.

### ScrapingBee vs ZenRows: which is better?

They overlap almost completely on capability. ZenRows markets harder on bypassing the most aggressive anti-bot systems and sells adjacent proxy and scraping-browser products. ScrapingBee is the steadier option: clearer documentation, billing only on successful requests, and published prices all the way through the enterprise range. Test both against your actual target list, since the only comparison that matters is cost per usable result on the sites you care about.

### Can I use ScrapingBee without knowing how to code?

Only indirectly. There is no visual scraper builder or dashboard for building jobs, so the direct product requires writing HTTP requests. The Make, n8n, and Zapier connectors let a no-code automation call the API, and the MCP server lets an AI agent do it, but anyone who wants to point at a page and click through a wizard should look at a tool like Octoparse instead.

### Does ScrapingBee handle CAPTCHAs and Cloudflare?

It handles a large share of anti-bot defenses through proxy rotation, browser fingerprint management, and stealth proxy mode, and most Cloudflare-protected pages return successfully with the right configuration. It is not marketed as a CAPTCHA-solving service, and the hardest targets still fail. Because failures are not billed, the practical way to find out is to run a sample of your real URLs through the free credits.

### Is web scraping with ScrapingBee legal?

The tool does not change the legal analysis. Collecting publicly available data is broadly permitted in many jurisdictions, but a site's terms of service, copyright, database rights, and privacy law (GDPR in particular, where personal data is involved) all still apply, and the answer differs by country and by dataset. Treat the API as infrastructure and take your own legal advice for anything involving personal data or a site whose terms you have accepted.

### How is ScrapingBee different from a proxy provider?

A proxy network sells IP addresses by the gigabyte and leaves the browser, retries, and evasion to you. ScrapingBee sells the completed request: proxies, headless Chrome, retries, and extraction bundled and priced per call. If your targets do not need rendering and your volume is high, raw proxies are cheaper. If a meaningful share need a real browser, the bundled API usually costs less than assembling the same stack yourself.

### What are the concurrency limits?

Concurrency is set by plan: 25 simultaneous requests on Hobby, 50 on Freelance, 100 on Startup, 200 on Business, 400 on Business+, and 500 to 900 across the enterprise steps. This is a separate constraint from credits, so a large backfill can be throughput-limited even on a plan with plenty of allowance left.

## Editorial verdict

ScrapingBee is the sensible default for a developer who needs web pages to come back reliably and does not want to run proxy contracts and a browser farm to get them. Its virtues are unglamorous and durable: a small API, honest documentation, billing only on successful responses, and published prices from $19.99 all the way to the top of the enterprise range, which almost nobody else in web data does. The costs it will impose on you are equally predictable. Credit multipliers mean the same code can cost fifteen times more depending on one boolean, credits expire monthly, and the product deliberately stops at fetching, so scheduling, storage, and orchestration remain yours. The Oxylabs acquisition in 2025 has so far been benign, but it does mean the roadmap now belongs to a proxy company rather than to two founders. Buy it as a component, size the plan against your real mix of rendered and stealth requests rather than the headline credit count, and keep a visual tool in mind if anyone non-technical ever needs to run a job.

---

Source: SaaSTracker (https://saastracker.org), an independent editorial project. This profile is compiled from public information, carries no peer reviews or paid placement, and was last reviewed 2026-08-23. Awards are judged on published criteria: https://saastracker.org/methodology