Best Rotating Proxy API Services Compared for 2026

Scraping targets that accept datacenter IPs, at steady volume? Buy a rotating gateway billed by concurrency, not by traffic. SparkProxy sells that: 100 threads for $75/mo on Starter up to 1000 threads for $440/mo on Plus, unlimited bandwidth, rotation per request or every 5 minutes, across 1M+ datacenter IPs in 80+ countries.

Targets that block datacenter ranges? You need residential exits, which SparkProxy does not sell. Decodo publishes self-serve per-GB residential pricing with a low entry point. Oxylabs is the pick if you need very long sticky sessions.

Heavily protected targets where your engineering time costs more than the requests? Stop buying IPs and buy outcomes. Zyte API has the lowest published per-request floor for simple sites. Bright Data Web Unlocker and Oxylabs Web Scraper API cost more per page but come from vendors with large enterprise footprints.

Want managed fetching without per-site difficulty pricing? A credit-based scraping API (SparkProxy's Scraping API, ScraperAPI) prices by what you ask for, not by how hard the site is.

No vendor here is best at everything. Each entry below says where it loses.

A rotating proxy API is an endpoint that hands each request, or each session, a different exit IP without you managing a proxy list. You point your HTTP client at one hostname and port, and the provider picks the address.

That is the whole concept, and our explainer on what a rotating proxy API is covers the mechanics, credential strings and session handling in depth. This post assumes you already know why you want rotation and are deciding whose rotation to pay for. If you are still at the "why rotate at all" stage, start with how proxy rotation works.

The vendors in this space sell one of two things, and the first buying decision is which one.

A rotating gateway (you buy IPs). You send normal proxied traffic to a host such as gateway.sparkproxy.io:11000. The provider rotates the exit address. Everything else is still yours: headers, TLS fingerprint, cookies, retries, CAPTCHA handling, JavaScript rendering. When a request is blocked, you pay for it (on per-GB plans) or you just lose the time (on flat plans), and your code decides what to do next.

An unblocker or scraping API (you buy responses). You send a URL to an HTTP API. The provider rotates IPs, retries, manages fingerprints, optionally renders the page, and returns the result. You pay per request or per successful response. Your code gets much simpler, and your cost per page goes up, often by an order of magnitude.

The rule of thumb that falls out: if your success rate through a plain rotating gateway is already high, a gateway is almost always cheaper. If it is low and you would need to build retry, fingerprint and rendering logic to fix it, price an unblocker before you build. You cannot make that call without knowing your success rate, so measure it first; our guide to proxy success rate shows how.

1. Billing unit. Per GB, per concurrent thread, per IP, per request, or per successful response. This choice moves the bill more than any per-unit discount.
2. What counts as billable. "Success" is defined by the vendor, and definitions differ. A 404 is a success on some APIs.
3. IP type. Datacenter, residential or mobile. Rotation does not fix an IP type the target distrusts.
4. Rotation controls. Per-request rotation, timed rotation, and how long a sticky session can hold one IP.
5. Concurrency ceiling. Throughput is set by requests in flight, not by the pool size on the marketing page.
6. A trial on your own targets. A free tier you can point at the real site beats any published success-rate figure, including ours.

Deliberately ignored: vendor uptime and success-rate claims. Every provider prints a number in the high nineties, and none are independently audited.

All competitor rows are from the vendor's own pages as of September 2026. Verify before buying.

SparkProxy (rotating datacenter gateway)

One hostname, three ports: gateway.sparkproxy.io on 11000 for HTTP/HTTPS, 11002 for sticky sessions, 13000 for SOCKS5. Rotation is random per request or automatic every 5 minutes, with USA and worldwide pools. Plans are Starter $75 (100 threads, 5 whitelist slots), Core $140 (250, 10), Boost $240 (500, 15) and Plus $440 (1000, 25), all 30 days and unlimited bandwidth. The fair usage policy sets a speed ceiling per plan, 25 Mbps on Starter up to 150 on Plus, which is a cap and not a promised rate.

Where it loses: it is datacenter only. If a target rejects hosting ASNs, rotation will not save you. The $75 floor is also poor value if you only send a few thousand requests a month.

Decodo (rotating residential gateway)

Decodo's residential page lists a 115M+ IP pool in 195+ locations, $4.00/GB pay as you go, and monthly plans from $3.75/GB on 3 GB down to $2.75/GB on 100 GB. It offers rotating and sticky sessions, with sticky sessions of up to 30 minutes. Self-serve signup and a 3-day, 100 MB trial make it easy to test.

Where it loses: per-GB billing charges for blocked responses and heavy pages alike. A 30-minute sticky cap is short for long logged-in workflows.

Oxylabs (rotating residential gateway)

Oxylabs lists a 175M+ residential pool across 195+ countries and sticky sessions that can hold one IP for up to 24 hours, the longest published session window in this comparison. Pricing is per GB across tiered plans, and the free trial is requested through a contact form rather than instant signup.

Where it loses: the trial is sales-assisted and its per-GB entry tiers sit above Decodo's. Check their residential page for the current rate card.

Webshare (per-proxy lists)

Webshare prices proxy servers per address, from $0.0299 per proxy at 100 proxies down to $0.0179 at 60,000+, with 10 free proxies and no card. You get a list and rotate across it, or use their rotating residential product billed per GB.

Where it loses: the per-proxy price is only half the bill, because bandwidth is selected as a separate tier. A per-IP list also means your code owns the rotation and ban logic.

Zyte API (unblocker)

Zyte bills per successful response and prices by how hard the website is, across five tiers from Simple to Advanced. Pay-as-you-go HTTP responses run from $0.13 per 1,000 (Simple) to $1.27 (Advanced), browser-rendered from $1.01 to $16.08. Monthly commitments of $100, $200 or $500 cut those rates. New accounts get $5 of credit for 30 days.

Where it loses: you do not pick the tier, the site does. The same million requests can cost ten times more on a protected target, and the browser-rendered Advanced tier is the most expensive published rate here.

Bright Data Web Unlocker (unblocker)

Web Unlocker is listed at $1.50 per 1,000 requests pay as you go, with success-based billing. The Scale plan is $499/mo including 383,000 requests, then $1.30 per 1,000. A free tier covers 5,000 requests a month, and new accounts get a first-deposit match up to $500.

Where it loses: on easy sites it is several times the price of a plain gateway or Zyte's simple tier. You pay for unblocking capability whether the target needed it or not.

Oxylabs Web Scraper API (scraping API)

Plans are Micro $49/mo (up to 98,000 results), Starter $99 (220,000) and Advanced $249 (622,500). Published per-1,000 rates without rendering are $0.40 to $0.50 for Amazon, $0.80 to $1.00 for Google and $0.95 to $1.15 for other sites; JavaScript rendering is $1.25 to $1.35. Their page says unsuccessful attempts are not charged, but 2xx and 4xx responses count as billable.

Where it loses: general-site pricing near $1 per 1,000 makes it expensive for simple pages, and the published plans top out below one million results.

ScraperAPI (scraping API)

ScraperAPI's documentation prices in credits: a normal request costs 1 credit, premium=true with render=true costs 25, and ultra_premium=true with render=true costs 75. They state they charge only for successful requests, defined as 200 and 404 responses, plus requests you cancel before 70 seconds.

Where it loses: the premium multipliers turn a 1-credit plan into a 25x or 75x plan on protected sites, so size plans on the multiplier you will actually need.

Almost every unblocker advertises success-based billing. Read the definition, because it varies in ways that matter.

The detail most comparisons miss: a 404 is a "success" on at least two of these APIs. If your URL list is stale, say 15% dead product pages, you pay for all of them. On a crawl that discovers URLs by guessing pagination or IDs, that share can be much higher. Clean your URL list before sending it to a per-success API.

The mirror image applies to gateways. On a per-GB plan, a block page is billable bytes. On a flat thread plan such as SparkProxy's, a blocked request costs time, not money. Neither is "success billing", but the flat plan has the same property for retries: they do not raise the bill.

This section is an illustrative model, not a benchmark. We did not run this workload against these vendors. The assumptions are chosen to be plausible so the pricing models can be compared; your numbers will differ.

Assumptions. One million plain HTML requests in a month. Average response 150 KB, so about 143 GB of transfer. Target accepts datacenter IPs. Zero failures, which flatters every per-GB and per-request model.

Now change one thing: every page needs JavaScript rendering.

• SparkProxy's Scraping API charges 5 credits per render, so 5,000,000 credits, which is the Scale plan at $599.
• Zyte's browser-rendered Simple tier is $1.01 per 1,000, so about $1,010, rising to about $4,020 on Moderate.
• Oxylabs Web Scraper API rendering is $1.25 to $1.35 per 1,000, about $1,250 to $1,350.
• A rotating gateway does not render at all. You run the headless browser, and a rendered page can weigh 2 MB or more, which turns 143 GB into roughly 1.9 TB on a per-GB plan.

Two lessons. First, on an easy target a flat gateway costs between 1.7x and 17x less than the unblocker rows in this model, which is why unblockers only make sense where the gateway fails. Second, per-difficulty pricing (Zyte) is the cheapest option on easy sites and one of the most volatile on hard ones. For more on how flat and metered pricing behave, see our breakdown of datacenter proxy pricing models.

Rotation is not one feature. Check these four things on any trial.

Per-request rotation actually rotates. Fire ten requests through the gateway at an IP echo service and count distinct addresses:

for i in $(seq 1 10); do
  curl -s -x http://USER:PASS@gateway.sparkproxy.io:11000 https://api.ipify.org
  echo
done | sort | uniq -c

Sticky sessions hold for as long as your flow needs. Published maximums in this comparison range from 30 minutes (Decodo) to 24 hours (Oxylabs). A checkout or login flow that takes 40 minutes breaks on the shorter one. For how sessions pin an address, read what a sticky session proxy is.

import requests

# Sticky port: keep one exit IP across a multi-step flow
STICKY = "http://USER:PASS@gateway.sparkproxy.io:11002"
s = requests.Session()
s.proxies = {"http": STICKY, "https": STICKY}

first = s.get("https://api.ipify.org", timeout=20).text
second = s.get("https://api.ipify.org", timeout=20).text
print("same IP across steps:", first == second)

Concurrency limits match your throughput plan. A thread plan caps requests in flight; a scraping API caps concurrent calls (SparkProxy's Scraping API allows 50 on Starter up to 400 on Scale). Size it with our guide to concurrent connections.

Cost is visible per request. On an API, you want the charge on every response so you can cost a job on 1,000 URLs before running a million. SparkProxy's Scraping API returns it in a header:

curl -G -i "https://scrape.sparkproxy.io/api/v1" \
  -H "X-API-Key: YOUR_API_KEY" \
  --data-urlencode "url=https://www.sparkproxy.io/" \
  --data-urlencode "render_js=false" \
  --data-urlencode "country_code=US" | grep -i "x-credits-used"

With render_js=false the call costs 1 credit instead of 5.

If you are torn between the top row and the bottom three, run the same 1,000 URLs through a gateway and through an API, and compare cost per successful record. Our comparison of scraping APIs vs self-managed proxies walks through the engineering cost side of that decision.