The short answer
- 1 Proxy-Seller Best all-round pick for scraping Visit site
- 2 Oxylabs.io Best for enterprise volume Visit site
- 3 Decodo (ex Smartproxy) Best targeting granularity Visit site
17 best proxies for web scraping, compared
| # | Provider | Fit | From | Readers | Protocols | Links |
|---|---|---|---|---|---|---|
| 1 | 4.6/5 | $0.30 / GB | 96% | HTTP, HTTPS, SOCKS5 | ||
|
A 47M+ residential pool across 220+ locations at $0.30/GB, with HTTP, HTTPS and SOCKS5 and a REST API — the broadest combination at the lowest published entry rate in this table. Good for this job
Watch out
|
||||||
| 2 | 4.5/5 | $0.40 / GB | 100% | — | ||
|
175M+ residential addresses across 195 locations with unit rates published at every tier, falling from $6/GB to $2.5/GB — the only network here you can budget at scale without talking to anyone. Good for this job
Watch out
|
||||||
| 3 | 4.4/5 | $0.08 / GB | 100% | HTTP(S), SOCKS5 | ||
|
115M+ IPs across 195+ locations with city, ZIP and ASN targeting and sticky sessions up to 30 minutes — the finest geographic control in this table. Good for this job
Watch out
|
||||||
| 4 | 4.3/5 | $0.02 / month | 100% | — | ||
|
A permanent free plan with 10 proxies and no card, then per-IP pricing from $0.0299 down to $0.0179 — the cheapest way to get a scraper working before spending anything. Good for this job
Watch out
|
||||||
| 5 | 4.3/5 | $0.35 / GB | 90% | HTTP, HTTPS, SOCKS5, HTTP/2 | ||
|
Unused gigabytes never expire on GB-based plans, with 30M+ IPs across 170+ countries and targeting down to ZIP, ASN and OS/TCP at $0.35/GB. Good for this job
Watch out
|
||||||
| 6 | 4.3/5 | $0.50 / GB | 88% | HTTP(S) | ||
|
100M+ IPs across 195+ countries at $0.50/GB, and the only provider here that publishes per-country IP counts rather than one global number. Good for this job
Watch out
|
||||||
| 7 | 4.2/5 | $0.90 / GB | 86% | HTTP, HTTPS, SOCKS5 | ||
|
The only provider here with third-party country-level tests on record: 96.8% in Mexico to 98.7% in Germany, with response times from 0.52s to 1.00s. Good for this job
Watch out
|
||||||
| 8 | 4.1/5 | $0.06 / GB | 100% | HTTP, HTTPS, SOCKS5 | ||
|
86M+ residential IPs across 200+ countries with country, state and city targeting, usage-based billing and static IP products on the same account. Good for this job
Watch out
|
||||||
| 9 | 4.1/5 | $2 / GB | 88% | SOCKS4, SOCKS5, HTTP, HTTPS | ||
|
Pay-per-GB with traffic that never expires and rolls into the next plan, unlimited concurrent connections, and targeting down to ZIP and ASN. Good for this job
Watch out
|
||||||
| 10 | 4.0/5 | $2.20 / GB | 88% | SOCKS5, HTTP, HTTPS | ||
|
Sticky sessions quoted up to 24 hours with unlimited concurrency, and residential, mobile and ISP under one dashboard — but at $2.20/GB, which adds up fast on volume. Good for this job
Watch out
|
||||||
| 11 | 3.9/5 | — | 81% | SOCKS5, HTTP, HTTPS | ||
|
80M+ residential IPs with rotating, static, datacenter, ISP and unlimited sold and billed as separate products, so one line's usage never eats another's budget. Good for this job
Watch out
|
||||||
| 12 | 3.9/5 | $0.63 / GB | 86% | SOCKS5, HTTP, HTTPS | ||
|
Rotating, static residential, static datacenter, mobile and ISP on one account from $0.63/GB, with 20% off at 96 GB and 30% at 120 GB published as open tiers. Good for this job
Watch out
|
||||||
| 13 | 3.9/5 | $2 / GB | 80% | HTTPS, SOCKS5 | ||
|
Residential and mobile on a single plan at a single price with only bandwidth metered, unlimited rotation and concurrency, and sticky sessions up to 7 days — but only 38+ countries. Good for this job
Watch out
|
||||||
| 14 | 3.8/5 | $5 / proxy | 85% | — | ||
|
Priced per proxy per month rather than per gigabyte, from $5, with volume discounts at 10, 50 and 100 proxies — predictable in a way per-GB billing is not. Good for this job
Watch out
|
||||||
| 15 | 3.6/5 | $2 | 82% | HTTP/S | ||
|
IPs the provider says are allocated to you alone, with no concurrency limit and rollover of unused bandwidth — but the published pricing contradicts itself in several places. Good for this job
Watch out
|
||||||
| 16 | 3.5/5 | ₽490 / day | 83% | SOCKS5, HTTP | ||
|
Mobile proxies rented by the day with unlimited traffic, which suits a two-week campaign far better than a long crawl — there is no large residential pool here. Good for this job
Watch out
|
||||||
| 17 | 3.2/5 | $1 | 82% | — | ||
|
Shared proxies from $1 with unlimited bandwidth and instant activation, which is useful for a one-off check and wrong for a crawl — the provider itself warns performance suffers on busy servers. Good for this job
Watch out
|
||||||
Ordered by how well each provider fits this job, from the figures the providers publish, our review scores and reader votes — see our methodology.
Why a scraper needs proxies at all
A scraper is a client that behaves nothing like a person. It asks for a hundred pages in the time a reader asks for one, it never loads an image it does not need, and it comes back tomorrow at the same speed. From the other side all of that is one address doing something no browser does, and the cheapest response is to stop answering it.
What you hit first is rarely a ban page. It is the quieter stuff: a 429 after a few hundred requests, a CAPTCHA on a page that served fine an hour ago, prices that come back in the wrong currency because the site decided where you are from, or an empty product grid that renders perfectly in your own browser. A pool of addresses turns one suspicious client into many ordinary-looking ones, and that is the whole trick.
It also fixes a problem that has nothing to do with blocking. A great deal of what people scrape is different in different places — search results, shipping costs, availability, ad creatives. Without a proxy in the country you care about you are not collecting the wrong data slowly, you are collecting the wrong data perfectly.
Residential, datacenter, ISP or mobile: which one a scraper wants
The four types are not quality tiers. They are different trade-offs, and picking by price alone is how scraping budgets get wasted twice.
Datacenter
Fast and cheap, and announced to the world as belonging to a hosting provider. For a site that does not care — documentation, open data, your own staging environment, most APIs — datacenter addresses are the correct answer and anything else is overpaying. For a retailer or a search engine they are the first thing filtered.
Rotating residential
Addresses that belong to real consumer connections, handed out from a pool. This is the default for anything commercially interesting, and it is billed by the gigabyte, which changes how you write the scraper: every retried image and every unnecessary page is now a line on an invoice.
ISP, also sold as static residential
A residential-looking address that stays yours. The point is not anonymity, it is continuity — a session that has to survive a login, a cart or a multi-step form cannot be handed a different address mid-way. Usually billed per IP per month rather than per gigabyte, which is far kinder to a scraper that moves a lot of data through few identities.
Mobile
Carrier addresses, shared by thousands of real phones behind the operator's NAT. That sharing is the value: blocking one is expensive for the site, so mobile addresses survive where others do not. They are also the most expensive per unit of traffic, so they belong on the handful of targets that actually need them rather than across a whole crawl.
In practice most serious scraping runs on two of these at once — datacenter or ISP for the volume that nobody defends, rotating residential for the parts that are defended, and a mobile pool kept in reserve for the two or three endpoints that refuse everything else.
How to set a proxy up in your scraper
Every provider in the table hands you the same two things: a gateway host with a port, and credentials. The differences are in how you ask for a country and how you hold a session, and both are usually encoded in the username rather than configured anywhere.
- Take the endpoint from the dashboard, not from a blog post. Gateway hostnames and ports change, and a stale one looks exactly like a dead proxy.
- Authenticate the way your environment allows. Username and password travels anywhere. IP whitelisting is cleaner but needs a fixed address at your end, which a laptop and most CI runners do not have.
- Put the country and the session in the username. The common shape is
user-country-us-session-abc123. Keep the session token stable while you need the same address and change it when you want a new one — that is your rotation control, and it is finer than any dashboard toggle. - Send it through your HTTP client properly. In Python that is a
proxiesdict onrequestsor aproxyargument inhttpx; in a headless browser it is a launch flag plus a credentials callback, because Chromium will not take a password in the URL. - Verify before you crawl. One request to an echo endpoint tells you the exit address and the country it is seen as. Our IP checker answers the same question in a browser, and the proxy checker will tell you whether a list of addresses is alive and how anonymous each one looks before you build anything on top of it.
- Then handle failure as normal. Retry with a fresh session rather than the same one, back off on 429 instead of hammering, and treat a CAPTCHA as a signal to slow down rather than a page to parse.
Free proxies versus paid for scraping
We publish a large free proxy list and re-check it around the clock, so we are in a decent position to say what it is and is not for.
A free proxy is shared with strangers, it is usually someone's misconfigured server rather than a donated one, and its life is measured in hours. For a crawl that needs a thousand clean requests in a row, that is not a cheap version of a paid pool — it is a different thing that happens to speak the same protocol. Anything with a login behind it, anything billed to a card, anything where a leaked request matters: use a paid pool, and prefer one that tells you where its addresses come from.
Where the free list earns its place is before you spend anything. It will tell you whether a proxy fixes your problem at all, whether the site you care about reacts to geography, and whether your code handles a dead address the way you think it does. That is worth an afternoon, and it costs nothing.
Is scraping with proxies legal?
Nothing here is legal advice, and the honest summary is that the proxy is almost never the interesting part.
Using a proxy is ordinary network plumbing — the same mechanism as a corporate gateway or a VPN. What carries risk is what you collect and what you do with it: personal data pulls in the GDPR and its equivalents regardless of how you fetched it, copyrighted material is still copyrighted when it arrives through a residential address, and a site's terms of service may prohibit automated access whether or not that prohibition is enforceable where you are. Volume matters too: a crawl that degrades a service is a different conversation from one that does not.
The practical version: scrape public pages rather than anything behind a login you agreed terms to
get, respect robots.txt and a sane rate even when nobody is forcing you to, keep personal
data out of it unless you have a lawful basis, and prefer a provider that states where its residential
addresses come from. Several in the table publish ISO or SOC audits, and when procurement asks that is
the paperwork they want.
Five mistakes that get a scraper blocked
- Rotating on every request by default. A session that changes address between page one and page two of the same flow is more suspicious than one that does not. Rotate per job, not per request.
- A perfect request. No
Accept-Language, no referrer, headers in an order no browser uses, and a residential IP in Berlin asking for everything in English. The address is clean and the request is not. - Paying residential rates for assets. Images, fonts and analytics beacons bill the same as the HTML you wanted. Blocking them in a headless browser routinely cuts per-GB spend by more than half.
- Treating every 403 as a bad proxy. Burning through a pool because of a bug in your own parser is an expensive way to find the bug.
- Buying before testing the one country that matters. Headline pool sizes are global. Depth in Germany tells you nothing about depth in Chile, and most providers in the table sell a small paid entry point specifically so you can find out.
What to look for
- Pool size and rotation
- A pool is consumed as a target bans its way through it. Published size matters less than depth in the countries you actually need, which is why we keep the per-country figures where a provider publishes them.
- Cost per gigabyte
- Scraping is billed by traffic, so the unit rate is the whole budget. Watch the entry tier rather than the enterprise one: the advertised low price is usually the latter.
- Concurrency and threads
- A crawl's speed is threads multiplied by latency. Providers that cap concurrent sessions, or charge for them separately, cost more than their per-GB rate suggests.
- Targeting granularity
- Country is rarely enough. City, ZIP and ASN targeting is what lets you reproduce what a particular shopper sees, and it is the difference between collecting prices and collecting the right prices.
- Session control
- Anything with a login, a cart or a multi-step form needs the same address for the whole flow. Sticky-session length is the figure to compare.
- What happens to unused traffic
- Gigabytes that expire at the end of the month quietly raise the real rate for anyone whose work comes in bursts. Rollover is worth more than a small discount.
Every provider, in order
1. Proxy-Seller — Best all-round pick for scraping
4.6 / 5 for this job · from $0.30 / GB
Proxy-Seller is the one we would point a new scraping project at first, and the reason is unglamorous: it is the cheapest credible per-gigabyte rate here, and it is attached to a pool deep enough to matter. The provider puts the residential network at 47M+ rotating addresses across 220+ locations, and sells six proxy types plus an enterprise track from the same account, so a crawl that starts on residential and later wants IPv6 or static addresses does not need a second vendor.
Two things make it easy to recommend for work that has to pass review. It states ISO/IEC 27001:2022 and ISO 9001:2015 certification along with GDPR, CCPA and ePrivacy compliance, and it offers DPA and SCC paperwork — which is exactly the list procurement asks for. The IPv6 pool of 1.2M+ addresses is also unusual at this price, and for targets that accept IPv6 it is the cheapest traffic you will find anywhere.
The gap is mobile: 20+ countries against 220+ residential locations is a much thinner footprint, so mobile-only targets are not its strength.
Key features
- 47M+ residential IPs across 220+ locations, as declared
- $0.30/GB entry rate — the lowest published here
- Six proxy types plus IPv6 under one account
- ISO 27001 and ISO 9001 certification, DPA and SCC available
- REST API, IP whitelisting and SDKs
Best suited for: a scraping project that wants one vendor for several proxy types and paperwork that survives procurement.
2. Oxylabs.io — Best for enterprise volume
4.5 / 5 for this job · from $0.40 / GB
Oxylabs is the heavyweight in this table and it prices like one. The provider claims 177M+ addresses in total, 175M+ of them residential, across 195 locations, and it publishes a unit rate for every product and every tier — residential falls from $6/GB on Starter to $2.5/GB at enterprise level. For a crawl whose volume you can forecast, that means you can model the whole year before contacting sales, which is rarer than it should be.
The compliance story is the strongest here: ISO/IEC 27001:2022 covering three named products and SOC 2 Type II covering two named endpoints. If your data collection has to withstand a legal review rather than just work, this is the shortlist of one.
It is a poor fit for experiments. The only free entry is $5 in welcome credits — under a gigabyte at Starter rates — the cheapest paid plan is $30/month, and enterprise starts at $2,500/month through sales. VAT may apply on the lower tiers, so the headline rate is not always the invoice.
Key features
- 175M+ residential IPs across 195 locations, as claimed
- Published unit rates at every tier, $6/GB down to $2.5/GB
- ISO/IEC 27001:2022 and SOC 2 Type II on named products
- Full product range including SERP and e-commerce endpoints
Best suited for: a funded, high-volume programme where audited sourcing and a forecastable unit rate matter more than the entry price.
3. Decodo (ex Smartproxy) — Best targeting granularity
4.4 / 5 for this job · from $0.08 / GB
If what you are collecting changes by neighbourhood rather than by country, Smartproxy is the pick. The provider puts the pool at 115M+ ethically-sourced addresses across 195+ locations and sells targeting down to city, ZIP code and ASN — not just country, which is where most networks stop. For price monitoring, local search results or ad verification, that is the difference between collecting prices and collecting the right prices.
The mechanics are scraper-friendly: HTTP(S) and SOCKS5 so older SOCKS5-only clients still connect, rotating and sticky sessions with sticky addresses held up to 30 minutes, and a stated 99.99% uptime. The $0.08/GB entry figure is the cheapest headline here after Webshare's per-IP pricing.
Read the trial terms before you lean on them. The 3-day free trial and the 14-day money-back option are mutually exclusive, refunds apply to a first purchase only and are capped at under 20% of a residential plan's traffic, and pay-as-you-go at $4/GB is wallet-only and cannot be combined with a subscription.
Key features
- 115M+ IPs across 195+ locations, per the provider
- City, ZIP code and ASN targeting
- Sticky sessions held up to 30 minutes
- HTTP(S) and SOCKS5
- 99.99% uptime, as stated
Best suited for: price monitoring, local SERP collection and anything where the answer depends on the neighbourhood, not the country.
4. Webshare.io — Best free start
4.3 / 5 for this job · from $0.02 / month
Webshare is where to prove your pipeline works. The free plan is permanent, gives you ten proxies and asks for no credit card, which makes it the only network here you can wire into a scraper on a Sunday afternoon at no cost. When the project is real, pricing is published per IP and falls from $0.0299 to $0.0179 as volume grows, with another 30% off for yearly billing — a very different shape of bill from per-gigabyte billing, and a much better one for crawls that move little data through many identities.
Three separate pools — datacenter, static residential (ISP) and rotating residential — mean the split-the-crawl strategy works inside one account: cheap addresses for the undefended majority, residential for the parts that fight back.
Two things are hard to verify before buying. The unlimited bandwidth claim carries an asterisk pointing at a Fair Usage Policy, and the provider's own pages give both 195 countries and 50+ locations, so the real footprint is difficult to pin down. The ISP carrier list names Comcast and AT&T and then says "& more", which is not much help if you need a specific carrier.
Key features
- Permanent free plan: 10 proxies, no card
- Per-IP pricing from $0.0299 down to $0.0179
- Datacenter, static residential and rotating residential pools
- 30% off on yearly billing
- 99.97% uptime, as stated
Best suited for: proving a scraper works before paying, and for crawls billed better per IP than per gigabyte.
5. ProxyEmpire — Best rollover bandwidth
4.3 / 5 for this job · from $0.35 / GB
ProxyEmpire's distinguishing feature is what happens to traffic you did not use: on GB-based plans it rolls over and never expires. That sounds like a detail until you run a crawl that comes in bursts, where expiring gigabytes quietly raise the rate you actually paid. Combined with no contracts and no monthly minimums — everything is prepaid — it is the friendliest billing here for uneven work.
The network is smaller than the giants at 30M+ residential addresses, but the targeting is not: country, region, city, ISP, ASN, OS/TCP fingerprint and ZIP, with mobile coverage stated at 170+ countries. Protocol support runs to HTTP, HTTPS, SOCKS5 and HTTP/2, and there is a public API with an endpoint generator.
The trial is narrower than it looks: the $1.97 entry excludes static ISP and datacenter, and unlimited residential only comes on request. Unlimited bandwidth is sold by the day or week from $200/day with no rollover, so the feature that makes the GB plans attractive does not apply there.
Key features
- Rollover bandwidth — unused GB never expire
- 30M+ residential IPs across 170+ countries
- Targeting by ZIP, ASN and OS/TCP fingerprint
- HTTP, HTTPS, SOCKS5 and HTTP/2
- No contracts, no monthly minimums
Best suited for: crawls that run in bursts, where expiring traffic costs more than a slightly higher unit rate.
6. Novada — Best published country depth
4.3 / 5 for this job · from $0.50 / GB
Novada does something we wish everyone did: it publishes country-level IP counts, for the US, Brazil, the UK, Canada and France. A global pool figure tells you almost nothing about depth where you are working, so a provider willing to break the number down is giving you something you can check a plan against. The headline is 100M+ residential addresses across 195+ countries with city-level targeting, at $0.50/GB.
Mobile is presented separately and specifically — 2M+ addresses across 120+ carriers — which is more useful than a single blended figure, and session windows are stated outright: 1 to 120 minutes on residential, 360-minute sticky sessions on rotating ISP. For a scraper that has to hold a login for an hour, those are the numbers that decide it.
Not everything has a price. Unlimited Proxies, IPv6 and Datasets are quoted through an account manager, the 99.99% success figure drops to 99% for rotating datacenter so the headline does not apply to every pool, and the free trial covers the Scraper API only — the general offer is a $10 credit, which is a different thing.
Key features
- 100M+ IPs across 195+ countries, per the provider
- Published per-country IP counts for five major markets
- Mobile stated separately: 2M+ IPs, 120+ carriers
- Session windows stated: 1-120 min residential, 360 min rotating ISP
- $0.50/GB entry rate
Best suited for: a crawl concentrated in a few named countries, where published depth beats a global headline.
7. FlyProxy — Best measured success rates
4.2 / 5 for this job · from $0.90 / GB
Most of the numbers on this page are self-reported. FlyProxy is the exception worth noting: a third-party listing recorded country-level success rates from 96.8% in Mexico to 98.7% in Germany, with response times between 0.52s and 1.00s. That is not our test, but it is somebody's, and it is more than almost anyone else here offers.
The network is described as over 100 million addresses by the provider and 69 million by that same third-party listing — a gap worth knowing about — covering 195+ countries and regions with state and city targeting. Unlimited concurrent sessions are confirmed from both sides, which matters more for crawl throughput than pool size does, and protocol support covers HTTP and SOCKS5 at a $0.90/GB entry rate.
Two hard edges: the provider states no refunds once products are sold, and there are no mobile proxies in the line-up at all. The trial position is also muddled — the third-party listing says there is none while the provider advertises a free 500M traffic offer through sales.
Key features
- Third-party tested: 96.8-98.7% success, 0.52-1.00s response
- 195+ countries with state and city targeting
- Unlimited concurrent sessions, confirmed by both sources
- HTTP and SOCKS5, 99.9% uptime as stated
Best suited for: throughput-heavy crawls in well-covered countries, where concurrency and measured success rates matter more than mobile coverage.
8. ProxySale.com — Best for mixing per-GB and static
4.1 / 5 for this job · from $0.06 / GB
Proxy-Sale covers the two billing shapes a scraper usually needs. The residential product is 86M+ addresses across 200+ countries with country, state and city targeting, billed on usage; alongside it are static IP products for the parts of the job that have to keep an identity. HTTP, HTTPS and SOCKS5 are all supported on residential, so older clients are not a problem, and a Scraper API sits on top for teams that would rather not maintain the fetching layer.
It is a solid, unexciting choice, and the unexciting part is deliberate: the stated 99.9% uptime, the success rate and the response times are all the provider's own with no third-party audit shown, so treat them as claims rather than measurements.
Two restrictions to check before buying: the service is not available in mainland China by the provider's own policy, and the free trial covers the Web Scraper API only, not the proxy network — so there is no free way to assess pool quality.
Key features
- 86M+ residential IPs across 200+ countries, as counted by the provider
- Country, state and city targeting
- HTTP, HTTPS and SOCKS5 on residential
- Static IP products and a Scraper API on the same account
Best suited for: a mixed workload that needs rotating traffic and a few fixed addresses without opening a second account.
9. GonzoProxy — Best pay-as-you-go
4.1 / 5 for this job · from $2 / GB
GonzoProxy is built for work that cannot be forecast. Traffic is bought by the gigabyte, never expires, and rolls over into the next plan, and every plan carries unlimited concurrent connections — so a crawl that needs forty threads for one night and none for a fortnight is not penalised. Targeting is unusually complete for a smaller network: country, city, state, continent, ZIP code and ASN.
Session control is generous. Rotating and sticky sessions are both offered, with sticky addresses claimed up to 72 hours — longer than anything else here — which suits anti-detect browser farms where a profile is supposed to look like the same person for days. Protocol support covers SOCKS4, SOCKS5, HTTP and HTTPS, and crypto funding is available if that matters.
The limits are practical rather than technical. Mobile covers Poland and Ukraine only as of writing, the dashboard lacks detailed per-request logs, and there is no browser extension or desktop app — everything runs in the web interface.
Key features
- Pay-per-GB, traffic never expires and rolls over
- Unlimited concurrent connections on every plan
- Sticky sessions claimed up to 72 hours
- Country, city, state, ZIP and ASN targeting
- SOCKS4, SOCKS5, HTTP and HTTPS
Best suited for: irregular workloads and anti-detect browser farms that need the same address for days at a time.
10. NodeMaven — Best long sticky sessions
4.0 / 5 for this job · from $2.20 / GB
NodeMaven's case for a scraper is session length. Sticky sessions are quoted at up to 24 hours with unlimited concurrent sessions, which covers the awkward jobs: a login that must survive a working day, a form filled across many steps, a geo-verification flow that breaks if the address moves. Residential, mobile and ISP proxies sit under one account and dashboard, and ISP plans carry unlimited traffic with a 99.9% uptime guarantee, which is the right shape for high-volume work through few identities.
Protocol support is complete — HTTP, HTTPS and SOCKS5, with UDP marketed for ISP — and stated uptime is 99.99%.
The per-gigabyte rate is the problem. At $2.20/GB this is one of the more expensive entries here, and high-volume scraping on a traffic model gets expensive quickly; the unlimited-traffic ISP plans are the way to use it at scale. There is no free trial — the $3.50, 750 MB entry is a paid demo, and a third-party write-up says the free-trial route is restricted to registered companies that pass KYC. Smaller locations are reported to be uneven in depth.
Key features
- Sticky sessions up to 24 hours, unlimited concurrency
- Residential, mobile and ISP on one dashboard
- ISP plans with unlimited traffic and a 99.9% uptime guarantee
- HTTP, HTTPS and SOCKS5, UDP on ISP
Best suited for: stateful crawls that must hold one address for hours, ideally on the unlimited-traffic ISP plans rather than per-GB.
11. Nsocks — Best separated product lines
3.9 / 5 for this job
Nsocks sells five things rather than one network with options: rotating, static, datacenter, ISP and unlimited proxies, each as its own product with its own billing. For a team running several crawls with separate budgets, that separation is genuinely useful — the datacenter work cannot quietly consume the residential allowance. The provider claims 80M+ real-user residential addresses, third parties report country, city, ISP and ASN targeting, and protocol support covers SOCKS5, HTTP and HTTPS at a stated 99.95% uptime.
The payment list is the broadest here, including crypto, Alipay, WeChat Pay and UPI, which matters more than it sounds if cards are awkward where you are.
Verifying any of it before you pay is the problem. There is no free trial, the provider states it does not support refunds on products sold, and its own home page quotes two different pool sizes — 80M+ in one place and 90M+ in another. Our review of it also has no published pricing table to compare, because the provider does not publish one.
Key features
- Five product lines billed separately
- 80M+ residential IPs claimed
- Country, city, ISP and ASN targeting reported by third parties
- SOCKS5, HTTP and HTTPS, 99.95% uptime as stated
- Crypto, Alipay, WeChat Pay and UPI accepted
Best suited for: teams running several crawls that need separate budgets per proxy type, and who can live without a trial.
12. Floppydata — Best open volume discounts
3.9 / 5 for this job · from $0.63 / GB
Floppydata publishes its volume discounts instead of hiding them behind sales: 20% off at 96 GB and 30% off at 120 GB, as open tiers. For a crawl sitting near one of those thresholds that is a concrete, checkable saving, and it is paired with an unusually complete line-up for the price — rotating, static residential, static datacenter, mobile and ISP, all on one account. Rotating traffic starts at $1.00/GB covering residential, mobile and datacenter together, and static datacenter addresses go from $0.79 per IP per month in bundles of 20, 50 or 100.
Protocol support is SOCKS5, HTTP and HTTPS, so nothing exotic is required at your end.
Support is the weak point. There is no phone line, so a live voice escalation is not available, documentation is thin for advanced setups such as custom rotation logic, and the 100 MB per-type trial is a connectivity check rather than a load test — enough to confirm the credentials work, not enough to judge the pool.
Key features
- 20% off at 96 GB and 30% at 120 GB, published openly
- Rotating, static residential, static datacenter, mobile and ISP on one account
- Static datacenter IPs from $0.79/IP/month in bundles
- SOCKS5, HTTP and HTTPS
Best suited for: mid-volume crawls near a discount threshold that want several proxy types without a second vendor.
13. AnyIP.io — Best residential and mobile on one plan
3.9 / 5 for this job · from $2 / GB
AnyIP's billing is the cleanest idea here: residential and mobile addresses come on one plan at one price, with only bandwidth metered. Rotation and concurrent requests are unlimited and not charged separately, so you are not doing arithmetic on threads. Sticky sessions run up to 7 days — the longest in this table — and filtering covers country, ISP, region and precise GPS coordinates, which is unusual and genuinely useful for anything location-sensitive.
Protocol support is HTTPS and SOCKS5.
Coverage is the catch, and it is a real one for scraping: 38+ countries is a short list next to the 195+ several competitors claim, so a global crawl will hit gaps. The pricing also needs reading twice — the advertised $2/GB is the Enterprise rate, while extra traffic costs $5/GB on Micro and $4/GB on Basic. Crypto payments sit outside the refund policy; the 14-day money-back guarantee covers card purchases only.
Key features
- Residential and mobile on one plan, one price, bandwidth-only metering
- Unlimited rotation and unlimited concurrent requests
- Sticky sessions up to 7 days
- Filtering by country, ISP, region and GPS coordinates
- HTTPS and SOCKS5
Best suited for: location-sensitive work in a handful of covered countries that needs residential and mobile without two contracts.
14. ASocks.com — Best per-proxy budgeting
3.8 / 5 for this job · from $5 / proxy
ASocks bills per proxy per month rather than per gigabyte, from $5, with discounts at 10 proxies (5%), 50 (10%) and 100 (15%). For a scraper whose traffic is hard to predict, that inverts the usual risk: the bill is fixed and the data is not. Three product lines — mobile, residential and corporate — sit on the same account, and the provider claims 150+ locations and 7M+ addresses.
That pool is small next to the 80-175M figures elsewhere on this page, which makes it a better fit for steady, moderate crawls than for broad ones that burn through addresses.
Treat the numbers with care: the provider's headline figures — 99.7% success, 150+ locations, 7M+ IPs, 100,000+ clients — are all self-reported and unaudited. Mobile is three times the price of the other lines at $15 per month against $5.
Key features
- Per-proxy monthly pricing from $5, not per-GB
- Volume discounts at 10, 50 and 100 proxies
- Mobile, residential and corporate lines on one account
- 150+ locations and 7M+ IPs, as claimed
Best suited for: steady crawls with unpredictable traffic, where a fixed monthly bill is worth more than a big pool.
15. Live Proxies — Best private IP allocation
3.6 / 5 for this job · from $2
Live Proxies sells something the rotating pools do not: private allocation, with addresses the provider says are unique to you rather than shared. For a crawl that keeps getting caught in someone else's reputation damage, that is a real argument. There is no concurrency limit — unlimited threads and unlimited requests, as stated — sticky sessions run 60 minutes on rotating residential and 30 on rotating mobile, and unused bandwidth rolls over to the standby plan.
The problem is working out what it costs. The home page prices static residential at $65.00/plan while the pricing page marks it permanently unavailable, and the home page's $45.00, $65.00 and $55.00 starting prices do not reconcile with the "B2C plans starting from $70" on the pricing page. That is not a reason to rule it out, but it is a reason to get a quote in writing.
The free trial goes to custom-plan users after an agreement; B2C buyers are told to buy a small plan instead.
Key features
- Private IP allocation, not shared, as stated
- No concurrency limit: unlimited threads and requests
- 60-minute sticky sessions on rotating residential
- Unused bandwidth rolls over to the standby plan
Best suited for: crawls that keep inheriting other people's bans, where a dedicated allocation is worth chasing a quote for.
16. MobileProxy.Space — Best for short mobile runs
3.5 / 5 for this job · from ₽490 / day
MobileProxy.Space is not built for general scraping and we would not pretend otherwise. What it does well is short mobile work: the minimum rental is one day, mobile plans bill by rental period with unlimited traffic rather than by the gigabyte, and the dashboard carries a proxy checker, port checks, a speed test and a browser extension. For a two-week job against a target that only tolerates carrier addresses, paying for time instead of traffic is the cheaper arrangement by a wide margin.
Protocol support is SOCKS5 and HTTP, and mobile plus dedicated datacenter proxies sit under one account. Crypto and local payment methods are accepted.
For a broad crawl it is the wrong tool: there is no large rotating residential pool to work through. Third-party write-ups flag the prices as relatively high, some customers have reported problems with geographical locations, and there are reports of laggy proxies and occasional late-night maintenance windows.
Key features
- One-day minimum rental
- Mobile plans billed by period with unlimited traffic
- Mobile and dedicated datacenter under one account
- SOCKS5 and HTTP, with a checker and speed test in the dashboard
Best suited for: short, mobile-only runs where unlimited traffic for a fixed period beats per-gigabyte billing.
17. ChangeMyIP.com — Cheapest way to test a single IP
3.2 / 5 for this job · from $1
ChangeMyIP is last here for scraping and the reason is in the product, not the price. It sells shared proxies — multiple users on the same address, sharing bandwidth — and the provider states plainly that performance can suffer on busy servers. For a crawl that depends on consistent throughput, a shared address with neighbours you cannot see is the opposite of what you want.
Where it earns a place is the one-off: instant activation, unlimited bandwidth on the plans it sells, 1Gb/s ports, multiple subnets and crypto payment, from $1. If you need a single working address this afternoon to check whether a page differs by country, it is the cheapest route in this table.
Read the restrictions first. Torrents and paypal.com are blocked at the proxy level on every product, and a $10 minimum order applies even if all you want is one $1.00 shared proxy.
Key features
- Shared proxies from $1, instant activation
- Unlimited bandwidth on the plans sold
- 1Gb/s ports and multiple subnets
- Crypto accepted
Best suited for: a single throwaway address for a quick check, not a crawl that needs steady throughput.
October’s pick
Proxy-Seller
Proxy-Seller sells residential, mobile, ISP, IPv4, IPv6 and datacentre proxies under one account, with published entry prices and ISO 27001 certification. The residential footprint is broad; the mobile one is much thinner.
See the current offerWeb scraping proxies: questions we get
How many proxies do I need for scraping?
With a rotating pool the question is traffic, not count — you buy gigabytes and the provider hands out addresses. Count only matters for static ISP or datacenter proxies, where a rough rule is one address per concurrent worker plus a spare, so an eight-thread crawler wants about ten.
Are datacenter proxies good enough for scraping?
For anything that is not defended, yes, and they are far cheaper per gigabyte. For retailers, search engines, social platforms and ticketing they are usually filtered on the first request, because the address announces itself as hosting.
What is the cheapest way to scrape at volume?
Split the crawl. Send the undefended majority through datacenter or ISP addresses, keep rotating residential for the parts that are actually blocked, and stop downloading images and trackers you never parse. That combination moves the bill more than switching provider does.
Do I need sticky sessions?
Only where state has to survive. Logins, carts, checkouts and multi-step forms need the same address throughout; a flat list of product pages does not and will rotate more cheaply.
Why do my proxies work and the data still come back wrong?
Usually geography or headers rather than the proxy. A German exit address asking for English content with no referrer and no Accept-Language still looks like automation, and plenty of sites serve a different page to a visitor they cannot place.
Can I scrape with free proxies?
For testing, yes — our own list is free and continuously re-checked, and it will tell you whether a proxy fixes your problem before you pay. For production, no: shared addresses with a lifetime measured in hours will not finish a long crawl, and nothing with a login should go through one.
What does rollover bandwidth actually save?
It matters when the work is uneven. If a month ends with twenty unused gigabytes and they expire, the effective rate you paid was higher than the one advertised. Several providers in the table let unused traffic carry forward instead.
Is scraping with a proxy against the law?
The proxy is ordinary network plumbing. The risk sits in what you collect — personal data, copyrighted material, anything behind a login whose terms you accepted — and in the load you put on a service. This is not legal advice; if the project is commercial, get some.
How do I check a proxy is actually working before a crawl?
Make one request to an echo endpoint and compare the address and country it reports with what you asked for. Our IP checker does it in a browser and our proxy checker will test a whole list for liveness, speed and anonymity level.
Some links on this page are affiliate links: if you buy through them we may earn a commission, at no cost to you. It does not change the order above — see our monetisation policy.