ZenRows’ $19 Build plan gets you 45,000 credits a month. That covers 45,000 simple pages but only 1,800 pages on a site protected by Cloudflare or DataDome. This review turns the November 2026 credit rules into real page counts, compares ZenRows with ScraperAPI, Scrapfly, Bright Data and Firecrawl, and tells you who should pay for it.
Key takeaways
- ZenRows is a web scraping API. You send a URL, and it deals with proxies, headless browsers, CAPTCHAs and anti-bot walls for you.
- Credit weights are fixed: 1 for a basic request, 5 with JavaScript, 10 with premium proxies and 25 with both.
- The free plan gives you 5,000 credits every month with no card, which works out to 200 fully protected pages.
- Failed or blocked requests cost nothing, but mode=auto can quietly move a domain up to the 25x rate.
- Review scores look good but rest on few reviews: G2 shows 5.0 from 18 reviews and Trustpilot 3.1 from 2.
Fact-checked
We checked these key facts against the sources named, on November 19, 2026:
- The ladder goes up to $2,999 for Scale S5 (45 million credits) (source: usagepricing.com)
- Without protection, 15,000 pages at 1 credit each is 15,000 credits (source: docs.zenrows.com)
- Both are right. $19 is the monthly price for Build B1. $16 is what it works out to per month on annual billing (source: zenrows.com)
- It is an established company with published docs, public pricing and open-source tools on GitHub (source: github.com)
ZenRows review: the short verdict
ZenRows is a good choice for developers who need to scrape protected websites and don’t want to run their own proxy pool and browser farm. Prices are public, and you are never charged for a failed request. On a site behind Cloudflare, though, your real cost is 25 times the headline rate. If your targets are mostly unprotected pages, cheaper tools or your own setup will beat it.
| Platform | Score | Number of reviews | Date checked |
|---|---|---|---|
| G2 | 5.0 / 5 | 18 | November 2026 |
| ToolRadar (aggregate of SourceForge, G2, Capterra) | 4.6 / 5 | 188 | September 17, 2026 |
| Trustpilot | 3.1 TrustScore | 2 | November 2026 |
Take these numbers with care. G2’s perfect score comes from just 18 reviews. Trustpilot’s 3.1 rests on two reviews, and its page notes that ZenRows hasn’t invited customers to leave reviews. The ToolRadar aggregate has the largest sample of the three, at 188 reviews, so it is the most useful number here.
Best for: Python or Node developers, data teams that scrape e-commerce or travel sites, and AI agent builders who want clean Markdown from websites.
Skip it if: you need a point-and-click desktop scraper, you scrape millions of unprotected pages a month, or you aren’t comfortable estimating usage-based costs.
What ZenRows actually does (in plain English)
Think of ZenRows as a courier for web pages. You give it an address (a URL). It works out how to get past the front door and brings the page back as HTML, parsed JSON or Markdown.
That front door is the hard part. Many large sites use anti-bot services such as Cloudflare, DataDome, PerimeterX (now part of HUMAN) and Akamai. These services check your IP address, your browser fingerprint and how the request behaves. A plain Python script running from a data-centre IP gets blocked, or gets a CAPTCHA, within a few requests.
ZenRows sends your request through rotating residential proxies in more than 190 countries. When a site needs JavaScript, it renders the page in a real headless browser. It also handles the fingerprinting and retries that anti-bot systems look for. You can request JSON, CSS-selected fields, or Markdown that is ready for a large language model.
Who it’s built for
ZenRows is a developer tool. You use it through an API key, code samples or an SDK. It suits people who already write scrapers in Python, Node.js or similar languages and have run into blocking. In 2026 it also targets AI builders. Markdown output and the MCP server let tools such as Claude Code and Cursor fetch live pages directly.
What it is not
It is not a no-code desktop scraper of the kind that records clicks and exports to Excel. It is also not database software. One ranking page wrongly describes ZenRows as a tool for pulling data out of SQL Server or IBM AS400 systems. It has nothing to do with databases and works only on public web pages.
The 2026 ZenRows lineup: Fetch, Extract, Crawl, Browser, Batch, MCP and CLI
ZenRows now sells seven products, and all of them draw on one shared pool of credits. The main change is that the old Universal Scraper API has been renamed Fetch. Older tutorials, Stack Overflow answers and GitHub snippets still use the old name.
| Product | What it does | When you’d use it |
|---|---|---|
| Fetch (formerly Universal Scraper API) | Returns one URL as HTML, JSON or Markdown | Most scraping jobs |
| Extract | Turns pages into structured data | You want fields, not raw HTML |
| Crawl | Follows links from a starting URL across a site | Collecting a whole catalogue or docs site |
| Browser (Scraping Browser) | A hosted browser you control from Playwright or Puppeteer | Logins-free flows that need clicks and scrolling |
| Batch | Sends many URLs in one job | Thousands of product or listing pages |
| MCP Server (@zenrows/mcp) | Gives AI coding tools a web scraping tool | Claude Code, Cursor, Claude Desktop, Windsurf |
| CLI (@zenrows/cli) | Scrapes from the terminal | Quick tests and shell scripts |
Fetch API and its key parameters
Fetch is the product most people mean when they say “ZenRows”. Its behaviour is set by query parameters, and each one changes your credit cost.
- js_render=true loads the page in a headless browser so that content built by JavaScript appears. It multiplies the cost by 5.
- premium_proxy=true sends the request through residential IPs instead of data-centre ones. It multiplies the cost by 10.
- Both together cost 25 credits. This is the usual setup for pages behind Cloudflare or DataDome.
- mode=auto (Adaptive Stealth Mode) starts with the cheapest setup and moves up to JavaScript rendering or premium proxies only when a request fails.
- response_type=markdown returns clean Markdown with headings, lists, links and tables kept. It adds no credits of its own, but you can’t combine it with the outputs parameter.
- proxy_country sets geotargeting, for example to see US prices or UK search results.
Scraping Browser with Playwright, Puppeteer and Selenium
Browser gives you a remote browser endpoint. Your existing Playwright or Puppeteer script connects to it instead of starting Chrome on your own machine. Your page actions stay the same, and ZenRows provides the proxies and fingerprinting. Browser sessions cost 5 credits per minute. A script that leaves a session idle keeps using credits until the session closes.
Batch for thousands of URLs
Batch takes a list of URLs and processes them asynchronously. You don’t have to manage concurrency and retries in your own loop. Use it for a nightly run over a product catalogue.
MCP Server and CLI for AI coding tools
The MCP server is published on npm as @zenrows/mcp. Setup guides cover Claude Code, Cursor, Claude Desktop and Windsurf. The CLI is @zenrows/cli on npm, and its GitHub repo was updated on November 9, 2026. Both read your key from an environment variable called ZENROWS_API_KEY, which is why that string turns up so often in searches and config files. The ZenRows GitHub organisation hosts these tools, SDKs and example code.
ZenRows pricing in November 2026, explained
ZenRows changed its pricing on August 4, 2026. Every product now draws on one shared “credits” currency, and plans come in rungs within each tier. The official docs list monthly prices. The pricing page and G2 show lower annual-billing equivalents.
| Plan / rung | Monthly price | Headline (annual) price | Credits per month | Concurrency |
|---|---|---|---|---|
| Free | $0 | $0 | 5,000 | 5 |
| Build B1 | $19 | about $16 | 45,000 | 20 |
| Build B2 | $29 | n/a | 80,000 | 20 |
| Build B3 | $39 | n/a | 120,000 | 20 |
| Launch L1 | $69 | $57 | 250,000 | 50 |
| Launch L2 | $99 | n/a | 500,000 | 50 |
| Launch L3 | $129 | n/a | 700,000 | 50 |
| Growth G1 | $199 | $165 | 1,200,000 | 100 |
| Scale S1 | $549 | $456 | 5,000,000 | 200 |
| Scale S5 | $2,999 | n/a | 45,000,000 | 200 |
| Enterprise | Custom | Custom | Custom | 400 to 1,000+ |
ZenRows publishes headline annual prices only for the entry rung of each tier. The Free plan’s concurrency of 5 comes from ZenRows’ own pricing page. The paid-tier limits come from a third-party pricing breakdown (ClawEngine), so check them in your dashboard.
Free plan: what 5,000 credits really gets you
The free plan is a recurring monthly allowance and needs no credit card. One competing review calls it a one-time trial, which is wrong. Unused free credits don’t roll over. At 5,000 credits you get 5,000 basic pages, 1,000 JavaScript pages, or 200 fully protected pages per month. That is enough to test your real targets before you pay.
Monthly vs annual prices: why $16 and $19 both appear
Searchers often ask why the pricing page says $16 while the docs say $19. Both are right. $19 is the monthly price for Build B1. $16 is what it works out to per month on annual billing. The same pattern holds across tiers: $57 vs $69, $165 vs $199 and $456 vs $549. That is about 17% off, close to “pay for 10 months, get 12”.
Annual vs monthly
Only commit to a year after the free plan and one paid month show your real credit use. The $3 a month saved on Build means little if you picked the wrong tier.
Top-ups, rollover and other line items
Build users can buy top-up packs when they run low. B1 and B2 top-ups add 11,000 or 13,000 credits for $6.50. B3 adds 43,000 credits for $19.50. Top-up credits roll over for one billing period, but your plan’s monthly credits do not.
The credit multiplier: what your money really buys
The price you see is in credits, and the cost of a page depends on how well protected the site is. Here are the docs’ fixed weights (1x, 5x, 10x, 25x) applied to each plan.
| Plan | Basic pages (1x) | JS pages (5x) | Premium proxy pages (10x) | Protected pages (25x) |
|---|---|---|---|---|
| Free | 5,000 | 1,000 | 500 | 200 |
| Build B1 ($19) | 45,000 | 9,000 | 4,500 | 1,800 |
| Build B3 ($39) | 120,000 | 24,000 | 12,000 | 4,800 |
| Launch L1 ($69) | 250,000 | 50,000 | 25,000 | 10,000 |
| Launch L3 ($129) | 700,000 | 140,000 | 70,000 | 28,000 |
| Growth G1 ($199) | 1,200,000 | 240,000 | 120,000 | 48,000 |
| Scale S1 ($549) | 5,000,000 | 1,000,000 | 500,000 | 200,000 |
| Scale S5 ($2,999) | 45,000,000 | 9,000,000 | 4,500,000 | 1,800,000 |
Here is the same data as cost per 1,000 pages. On Build B1, basic pages cost about $0.42 per 1,000 and protected pages about $10.56 per 1,000. On Growth G1, protected pages fall to about $4.15 per 1,000. On Scale S1 they fall to about $2.75. Volume discounts are real, but the 25x weight affects your budget far more than the tier you choose.
Did you know?
The ZenRows docs say only successful requests are charged. A request that gets blocked or fails uses no credits, so retries on a hard site don’t eat your balance.
Worked example: scraping a Cloudflare-protected store
Say you track prices on 500 product pages a day from a store behind Cloudflare, which comes to about 15,000 pages a month.
- Without protection, 15,000 pages at 1 credit each is 15,000 credits. Build B1 covers that easily.
- With JavaScript rendering and premium proxies, each page costs 25 credits. That makes 375,000 credits.
- 375,000 credits is more than Launch L1 (250,000) but fits Launch L2 (500,000) at $99 a month.
The site’s protection takes you from a $19 plan to a $99 plan. This also explains a common complaint: “Why did a Cloudflare-protected site eat my whole monthly allowance in one afternoon?” At 25 credits a page, Build B1 runs out after 1,800 pages. With 20 requests running at once, a crawler gets through that in a few hours.
The mode=auto trap
mode=auto is convenient because it chooses the setup for you. It starts cheap and moves up only when a request fails, and you are billed only for the setup that succeeds. On a domain that always needs the full setup, though, every page ends up billed at 25x. If you didn’t plan for that, the bill can surprise you.
Check credits per page
Start with plain requests and add js_render and premium_proxy only for URLs that fail. If you use mode=auto, check the dashboard after the first few hundred requests to see which rate each domain settled on.
Hidden costs: bandwidth and browser minutes
Two other items draw on the same credit pool:
- Residential proxy bandwidth: 25,000 credits per GB. B1’s 45,000 credits cover about 1.8 GB.
- Browser sessions: 5 credits per minute. B1 covers about 9,000 minutes, or 150 hours, of session time.
Did you know?
According to the ZenRows pricing docs, 1 GB of residential proxy traffic costs 25,000 credits. That is more than half of Build B1’s monthly allowance.
Does it really get past Cloudflare?
ZenRows gets past most common anti-bot setups when you turn on both JavaScript rendering and premium proxies. The exact success rate is disputed, and the only number you should trust is the one you measure on your own targets.
The claims come from parties with something to sell:
- ZenRows (the vendor) puts its success rate at 99% in its own 2026 benchmark. This is a vendor claim, not independent data.
- Scrapfly (a direct competitor) cites a Scrapeway benchmark that puts ZenRows at 58%. It is third-party data, but a rival is promoting it, and benchmark results change as anti-bot vendors update their systems.
Is 99% real or just marketing? It is best read as a best case on targets ZenRows has tuned for. Success rates differ a lot between sites, and they can drop overnight when Cloudflare, DataDome or Akamai change their detection.
How to run the test on your own URLs with the free plan
- List 20 to 50 URLs you actually need. Mix unprotected pages with pages behind Cloudflare, DataDome or PerimeterX.
- Run each one as a plain request first, then with js_render, then with js_render plus premium_proxy.
- Note the HTTP status, whether the content you need is in the response, and the response time.
- Divide the credits used by the number of successful pages. That is your real cost per page.
- Multiply that number by your monthly page volume to find the right plan.
That small test tells you more than any review score or vendor benchmark. It also answers “Which settings do I need for a protected page?” For most Cloudflare and DataDome pages you will need both settings. Some sites only need one, though, and those save you 15 to 20 credits per page.
Setting it up: your first request in 5 minutes
You need only an API key and a terminal to make your first request. Here is the quickest route.
Make your first request
- Sign up at zenrows.com. The free plan doesn’t need a card.
- Copy your API key from the dashboard and save it as the environment variable ZENROWS_API_KEY.
- Run a plain request: curl “https://api.zenrows.com/v1/?apikey=$ZENROWS_API_KEY&url=https://example.com”
- If the page is blocked or empty, add &js_render=true&premium_proxy=true to the same URL.
- Add &response_type=markdown if you’re sending the result to an LLM.
- Check the dashboard’s usage view to see how many credits the request used.
In Python you can build the same request with the requests library, or use ZenRows’ official SDK. For AI coding tools, register @zenrows/mcp as an MCP server in Claude Code, Cursor or Windsurf and pass ZENROWS_API_KEY in its environment settings. The assistant can then fetch live pages when you ask it to.
Keep your key private
Your API key spends your credits. Keep it in an environment variable or secrets manager, and never commit it to a public GitHub repo.
The dashboard’s request builder generates request code from the options you tick. If its live preview hangs on a heavy page, you don’t need to keep reloading it. Copy the generated code, run it from your own script, and add a wait parameter so slow JavaScript has time to load.
Pros and cons
ZenRows’ strengths are in billing transparency and anti-bot handling. Its weaknesses all come from the multiplier.
Pros
- Public, detailed pricing with fixed credit weights in the docs.
- Failed or blocked requests are free.
- A recurring 5,000-credit free plan with no card required.
- One credit pool covers Fetch, Extract, Crawl, Browser and Batch.
- Markdown output and an official MCP server for AI workflows.
- Residential proxies in 190+ countries, with geotargeting.
- Low lock-in: Fetch is simple URL-in, HTML-out, and Browser works with standard Playwright and Puppeteer code.
Cons
- Protected pages cost 25x, so Build B1 covers just 1,800 of them.
- mode=auto can raise costs without you noticing.
- Bandwidth at 25,000 credits per GB adds up quickly.
- Plan credits don’t roll over. Only Build top-ups carry over, and only for one period.
- Review samples are small, so the high scores tell you less than they seem to.
- It’s a developer product, and people who don’t code will struggle.
What about lock-in if you later move to your own setup? It is low if you keep your parsing code separate from the fetch call. Switching then means changing one function, not rewriting the scraper.
ZenRows vs ScraperAPI, Scrapfly, Bright Data and Firecrawl
These tools overlap a lot, but each charges differently and is good at different jobs. Prices are entry points as published in late 2026. Each tool uses its own credit multipliers, so compare cost per successful page on your own URLs, not credits per dollar.
| Tool | Free tier | Pricing model | Anti-bot strength | Standout feature | Best for |
|---|---|---|---|---|---|
| ZenRows | 5,000 credits/month | Credits, from $19/mo (45K) | Strong with JS + premium proxy | No charge for failures, MCP + CLI | Developers on protected sites |
| ScraperAPI | Trial credits | Credits, Hobby $49/mo (100K) | Good | Simple API, structured endpoints | Teams on mid-protection sites |
| Scrapfly | 1,000 credits | Credits, Discovery $30/mo (200K) | Strong | Detailed anti-bot controls | Engineers who want fine control |
| Bright Data | Limited free tier | Web Unlocker about $1.50 per 1K successful requests (pay as you go) | Very strong | Huge proxy network, datasets | Enterprise volume |
| Firecrawl | 1,000 credits/month | Credits, Hobby from $16/mo annual | Moderate | LLM-ready crawling, open source | AI and RAG pipelines |
Pick ZenRows if you scrape sites that block you and want transparent rules and free retries.
Pick ScraperAPI if your targets are lightly protected and you want a larger base allowance. Note that its credit costs also rise on hard domains.
Pick Scrapfly if you want granular control over anti-bot behaviour and detailed logs.
Pick Bright Data if you work at enterprise scale and need flat per-request unlocking or ready-made datasets.
Pick Firecrawl if your main goal is turning websites into Markdown for an LLM, and anti-bot protection matters less.
Zyte and ScrapingBee are also worth adding to a shortlist. Zyte prices by how hard each site is to scrape, and ScrapingBee has a credit system similar to ZenRows’.
Is it cheaper to buy residential proxies yourself than to pay the 25x rate? Per gigabyte, often yes. But raw proxies don’t render JavaScript, solve challenges or fix browser fingerprints. On Cloudflare-heavy sites, you would be rebuilding most of what ZenRows does. If you would rather not build any of this in-house, hiring one of the web development companies that specialise in data pipelines is another option.
Is web scraping with ZenRows legal?
Scraping publicly available data is generally lawful in the US. The details depend on what you collect and how you use it. This is general information, not legal advice.
- Public vs logged-in data: US courts have generally treated scraping of public pages more kindly than scraping behind a login. Two examples are hiQ Labs v. LinkedIn and the 2024 Meta v. Bright Data ruling on logged-out scraping. ZenRows works on public pages, so don’t use it to get past logins you aren’t allowed to use.
- Site terms: A site’s terms of service can still create contract risk, especially if you have accepted them by creating an account.
- robots.txt: It isn’t law, but following it is good practice and shows good faith.
- Personal data: Collecting names, emails or profiles can bring in the GDPR in Europe and the CCPA in California, whatever tool you use.
Get advice from a lawyer before running a commercial project that collects personal data or scrapes a competitor at scale.
Who should buy ZenRows (and who shouldn’t)
Your decision comes down to the share of your target pages that need the 25x setup.
- Buy Build B1 or B3 if you scrape a few thousand protected pages a month for price tracking, research or lead lists.
- Buy Launch or Growth if you run daily jobs on protected e-commerce, travel or real estate sites. At those tiers, protected pages cost about $4 to $7 per 1,000.
- Use the free plan only if you’re building an AI agent prototype that needs occasional live pages through MCP.
- Skip ZenRows if most of your targets are unprotected. Plain HTTP requests or a cheaper API will cost far less.
- Skip ZenRows if you need tens of millions of protected pages. At that point Enterprise quotes, Bright Data or an in-house setup are worth comparing.
The next step costs nothing. Run 20 to 50 of your real URLs on the free plan and work out your credits per successful page. Then match that number to the pages-per-plan table above.
FAQ
Is ZenRows worth it?
It’s worth it if your targets block normal scrapers and your time costs more than the 25x multiplier. If your pages are unprotected, it usually isn’t.
Is ZenRows free?
Yes. There is a free plan with 5,000 credits a month, 5 concurrent requests and no credit card. Unused credits don’t roll over.
How much does ZenRows cost per month?
Paid plans start at $19 a month for Build B1 (45,000 credits), or about $16 a month on annual billing. The ladder goes up to $2,999 for Scale S5 (45 million credits). Enterprise pricing is custom.
How do ZenRows credits work?
Every request uses credits based on fixed weights: 1 for basic, 5 with JavaScript rendering, 10 with premium proxies and 25 with both. Bandwidth costs 25,000 credits per GB, and browser time costs 5 credits per minute.
Does ZenRows bypass Cloudflare?
On most Cloudflare-protected pages it does, with js_render and premium_proxy turned on. Its own 2026 benchmark claims a 99% success rate, which conflicts with a 58% figure cited by competitor Scrapfly, so test your own URLs.
What is the ZenRows Fetch API?
Fetch is the new name for the Universal Scraper API. It is the main endpoint that takes a URL and returns the page as HTML, JSON or Markdown.
Is ZenRows better than ScraperAPI?
ZenRows has the edge on protected sites and on clear pricing rules. ScraperAPI’s $49 Hobby plan gives you a bigger base allowance for lightly protected targets. Run the same URLs through both trials and compare.
Does ZenRows charge for failed requests?
No. ZenRows’ docs say only successful requests are charged. Blocked or failed requests use no credits.
Can I use ZenRows with Python, Playwright or Puppeteer?
Yes. Python can call the Fetch API directly or through the official SDK. Playwright and Puppeteer connect to the Scraping Browser endpoint.
Is ZenRows legit and safe?
Yes. It is an established company with published docs, public pricing and open-source tools on GitHub. Keep your API key private, and download the MCP server and CLI only from the official @zenrows packages on npm.












