Aproxy is a Hong Kong-operated proxy provider offering rotating residential, unlimited, long-acting ISP, static residential and datacenter proxies, plus scraping and SERP APIs, priced from $0.18 per IP or $0.50 per GB.
Used this tool? Rate it
Used this tool? Rate it
Aproxy is a commercial proxy network that rents out IP addresses. When you send a web request through it, the request leaves from someone else's connection rather than yours, so the destination site sees that other address instead of your own. The company sells this capability in five distinct forms — rotating residential proxies, unlimited residential proxies, long-acting ISP proxies, static residential proxies, and static datacenter proxies — and layers four data-retrieval APIs on top of the same infrastructure.
It is worth being precise about the category, because Aproxy is listed in a directory that mostly covers AI software and this product is not that. There is no model, no inference, no generation, and no assistant. Aproxy is network plumbing: a pool of exit addresses plus the authentication, routing, and billing needed to sell access to them. Some of the workloads it serves do feed AI systems — the company runs a dedicated landing page for AI training data collection — but the product itself is infrastructure, not intelligence. Read the rest of this page with that framing in mind, and treat any claim that a proxy service is "AI-powered" with the skepticism it deserves.
The operating entity is disclosed as Smart Innovation Technology Limited, named in the company's anti-money-laundering policy, which commits to compliance with the Bank Secrecy Act alongside Hong Kong and international AML regulations. That jurisdiction is independently corroborated: the company's Trustpilot profile is a claimed profile, verified as claimed in March 2025, and lists the contact location as Hong Kong. Two separate sources pointing to the same jurisdiction is a meaningful signal in a category where shell operations are common.
The headline network figures the company publishes are 70 million or more residential IPs across 195 countries, with a stated 99.9% uptime. Those numbers come from Aproxy's own marketing and have not been independently audited — a point developed at length in the limitations section below, and one you should hold onto before treating any of them as a procurement input.
In most software categories, you evaluate a product by asking whether it works. In the residential proxy category you have to ask a prior question: where do the IP addresses come from? The answer determines whether using the service is merely a commercial decision or an ethical one.
A residential proxy works because real consumer devices — phones, laptops, home routers — agree to forward traffic on behalf of paying customers. The reputable version of this arrangement involves informed consent and compensation: a user installs an app, is told in clear terms that their spare bandwidth will be resold, and receives something in return. The disreputable version involves bundling that consent into an unread dialog inside a free VPN or flashlight app, or dispensing with consent altogether by compromising devices. In the worst cases, residential proxy pools have been assembled directly from malware-infected machines, meaning the customer's traffic exits through a device whose owner has no idea it is happening and did not agree to it.
This is not a hypothetical concern. Proxy networks have been the subject of enforcement actions and security-vendor takedowns precisely because their pools overlapped with compromised devices. When you buy residential proxy bandwidth, you inherit whatever provenance sits behind it.
Aproxy's public statement on this is a single sentence on its residential proxy product page: the pool is described as ethically sourced from carefully selected partners that uphold business ethics and strict compliance, with a majority of the network composed of what the company calls Tier A model proxies. The phrase "70M+ Compliant Source IPs" appears on the homepage.
Read that carefully, because what it does not say matters as much as what it does. It does not name the partners. It does not describe the consent mechanism through which end users agree to have their bandwidth resold. It does not define what "Tier A model" means, and that term is not an industry standard with an agreed definition. It does not disclose which SDKs or applications the IPs are recruited through. It confirms that the IPs are acquired from third-party suppliers rather than built directly, which makes the provenance question a question about those suppliers — and they are unnamed.
To be fair and precise: no evidence was found that Aproxy's pool overlaps with any known malicious network, and no security vendor or researcher has published a finding against it. The honest characterisation is that the sourcing claim is unverified rather than disproven. There is a real difference between those two states, and it would be as wrong to imply wrongdoing as it would be to accept the claim at face value. What can be said definitively is that the disclosure is thinner than what a buyer with a compliance obligation would need, and that anyone in that position should ask for the supplier chain in writing before purchasing.
Rotating residential proxies. The flagship product, drawing on the 70 million-plus residential pool across 195 countries. Each request or session can exit from a different address, which is the standard approach for large-scale collection where per-address request volume needs to stay low. Listed from $0.50 per GB, though the homepage banner simultaneously shows $0.8/GB — a discrepancy discussed under pricing.
Unlimited residential proxies. Sold on a time basis rather than metered bandwidth, covering unlimited use of graded residential proxies with randomly assigned countries, plus unlimited sub-accounts and whitelist entries. This inverts the usual economics: for workloads with unpredictable or very high bandwidth needs, a flat time-based rate removes the risk of a runaway metered bill. The trade-off is the randomly assigned countries — you give up precise geographic targeting in exchange for volume.
Long-acting ISP proxies. The most technically distinctive offering. These support sessions of up to 12 hours with unlimited connections and threads, listed from $0.27 per GB. Session persistence is what separates this from rotating proxies: any workflow requiring a logged-in state — account management, multi-step forms, anything with a session cookie — breaks when the exit address changes mid-flow. A twelve-hour window covers a full working session.
Static residential proxies. Dedicated addresses billed per IP rather than per gigabyte, from $0.27 per IP. Suited to work where the same address must be presented consistently over time.
Static datacenter proxies. The cheapest tier at $0.18 per IP. Datacenter addresses are faster and more stable than residential ones but are also far easier for target sites to identify and block, since they resolve to hosting providers rather than consumer ISPs. This is the correct choice for targets that do not scrutinise IP provenance, and the wrong one for those that do.
Data APIs. Four products sit above the proxy layer. The Web Scraper API is listed from $0.39 per 1,000 results and covers more than 100 domains, with Google, Bing, YouTube, DuckDuckGo, Instagram, Facebook, Amazon and Walmart named explicitly. The SERP API is listed at $0.75 per 1,000 results, the Video Data API at $0.06 per GB, and a Universal Scraper API is also offered. These shift the billing unit from bandwidth to results, which is generally the better model when you care about extracted data rather than bytes moved, and it moves the burden of proxy rotation and retry logic onto the vendor.
The Web Scraper API and SERP API are both marked for free trial on the pricing display, though the site renders this as "Free Trail" — a spelling error that persists across the interface, and a small but real indicator of the editorial care applied to the site.
Data and market-research teams running price monitoring, competitor tracking, or catalogue collection at a scale where a single address would be rate-limited within minutes.
SEO practitioners verifying how results render across geographies. Search results are heavily localised, so checking rankings from the market you actually serve requires an exit address in that market — the SERP API addresses this directly.
Ad-verification and brand-protection functions confirming that campaigns render as booked in target regions and detecting counterfeit listings or unauthorised distribution.
Multi-account social media operators, which is where the long-acting ISP tier and the documented BitBrowser integration matter, since session persistence and a consistent fingerprint are both prerequisites.
AI and ML data teams collecting training corpora at volume — the company maintains a dedicated page for this use case.
Who should look elsewhere: anyone with a strict vendor due-diligence process that requires documented IP provenance, since the supplier chain is not disclosed. Anyone needing an independently benchmarked success rate for a contractual SLA, since no such benchmark exists. And anyone operating primarily from mainland China, given the explicit geographic restriction described below.
Access is through standard proxy endpoints over HTTP and SOCKS5, which is the broadest possible compatibility surface — anything that accepts a proxy configuration will work, from curl to a headless browser cluster to commercial scraping software.
A web dashboard handles usage statistics, sub-user creation, and whitelist management. A Public API is offered for programmatic proxy management, letting resource administration be scripted rather than clicked. The help centre is organised into basic knowledge, proxy types, system integration covering operating system and browser configuration, and API integration for the scraper products.
Named integrations are modest but relevant to the stated use cases: BitBrowser for anti-detect browsing and V2Ray for proxy tooling. Free utilities include a proxy checker, a country code lookup, and a free proxy list. Support runs through support@aproxy.com and a 24/7 live chat.
One structural observation: the site publishes localised sitemaps in eight languages — English, Simplified Chinese, Vietnamese, German, Indonesian, Russian, Hindi and Portuguese — with 682 URLs in the English tree alone, including per-country landing pages for the full ISO country list. This is a sustained content operation, not a single-page front. It says nothing about service quality, but it does argue against the site being an abandoned shell.
Published entry prices, taken directly from the site:
| Product | Listed entry price | Billing unit |
|---|---|---|
| Residential proxy | $0.50/GB (banner also shows $0.8/GB) | Bandwidth |
| Unlimited residential | $67.87/day | Time |
| Long-acting ISP | $0.27/GB | Bandwidth |
| Static residential | $0.27/IP | Per address |
| Static datacenter | $0.18/IP | Per address |
| Web Scraper API | $0.39/1K results | Results |
| SERP API | $0.75/1K results | Results |
| Video Data API | $0.06/GB | Bandwidth |
Three caveats belong with this table, and they are not decoration.
First, these are entry prices — "starting at" figures that typically correspond to the largest commitment tier. The effective rate on a small first purchase will be higher, and the site does not publish a full tier ladder without an account. If you need the actual rate for your volume, you will have to register to see it.
Second, the figures are internally inconsistent in at least two places. The residential product page shows $0.50/GB in one position and $0.8/GB in a banner on the same page. The unlimited residential page renders its rate as both $67.87/Day and $67.87/Hour — a difference of twenty-four times, which cannot be a rounding matter and is plainly an error in one of the two. Which one is correct is not determinable from the public site. Confirm the billing unit in writing before purchasing a time-based plan; this is exactly the kind of ambiguity that produces a disputed invoice.
Third, third-party listings carry different numbers again. A promotional article on the DiCloak blog lists $0.80/GB residential, $0.28/IP static residential, $68.79/day unlimited and $0.66/GB long-acting ISP — every figure differing from the current site. The most likely explanation is that prices moved between publication dates rather than anything improper, but it demonstrates that quoted rates for this service drift, and that no third-party figure should be trusted over the live site. That article, it should be noted, was published by a company selling a competing anti-detect browser and reads as promotional placement rather than independent testing.
Proxy services generate refund disputes at an unusually high rate, so the specifics here deserve attention. Aproxy publishes a dedicated refund policy, and it is more detailed than most in this category.
Refunds are accepted where the customer cannot connect after purchase or the supplied IPs are completely unusable and the problem is unresolved within 48 hours; where a systemic malfunction disrupts service for more than 24 consecutive hours; where a duplicate payment occurred; and in other cases the company evaluates as reasonable. Requests go to support@aproxy.com with the order number and supporting documentation, drawing a response within one to two business days and processing within three to seven business days back to the original payment method.
The exclusions are where the real risk sits. No refund is available for user configuration errors, for conflicts with the customer's network environment or third-party software, for orders with significant traffic usage or used for more than 48 hours absent severe quality problems, or for free trials and promotional gift packages. The company also reserves the right of final interpretation of the policy and may decline requests it judges to be abuse.
The practical reading is that the 48-hour window is the binding constraint. Whether a given failure counts as "the provided IPs are completely unusable" or as a "user operation error" is a judgement the vendor makes, under a policy where it holds final interpretation. Test the service hard and immediately after purchase — not next week.
Note also a separate and slightly different formulation in the terms of service, which states that fees paid are generally non-refundable unless required by applicable law or the refund policy, with special circumstances evaluated case by case. The refund policy is the more specific and more customer-favourable document; where the two differ, the dedicated refund policy is the operative one, but a buyer should be aware that two formulations exist.
No independent performance verification exists. This is the single most important caveat on this page. Proxyway, the industry's most established independent benchmarking outlet, publishes annual tested rankings of residential proxy providers and does not include Aproxy in them. The 99.9% uptime figure, the 70 million IP count and the 195-country coverage are all vendor self-reported and have not been audited by any third party identified during research. They are not disproven — they are simply unverified, and should not be entered into a procurement comparison as though they were measured.
Third-party review coverage is statistically meaningless. The Trustpilot profile carries a 3.3 score from exactly two reviews, split fifty-fifty between one five-star and one one-star rating, with zero reviews in the past twelve months. Two reviews cannot support any conclusion whatsoever about service quality. A G2 seller page exists and search summaries indicate a high rating, but that page returned HTTP 403 through both direct fetch and a real-browser fallback, so its contents could not be read and verified first-hand and are not relied on here. Community discussion is likewise absent: targeted searches for complaints, disputes or provenance concerns returned nothing about this provider in either direction.
Mainland China is excluded. The FAQ states plainly that, for policy reasons, the proxy must be used in a non-mainland-China internet environment. This is a hard operational constraint, not a performance note.
IP provenance is asserted but not documented. Covered fully above. The suppliers are unnamed and the consent mechanism is undescribed.
The ISO/IEC 27001:2017 badge deserves a second look. The site displays this as a certified product. ISO/IEC 27001 is an information-security management standard, and its widely adopted revisions are the 2013 and 2022 editions; a 2017 designation typically refers to a national or regional adoption of the 2013 text rather than a distinct international edition. More fundamentally, ISO 27001 certifies an organisation's security management processes — it says nothing about IP sourcing ethics, uptime, or success rate. It is not a substitute for the provenance disclosure that is missing.
Legal documentation is thin on entity detail. The terms of service names no legal entity and specifies governing law only as "the laws of our location" with disputes going to "the competent court in our location" — a formulation that leaves the customer unable to determine, from the terms alone, which courts have jurisdiction. Only the AML page names Smart Innovation Technology Limited. No company registration number is published anywhere on the site.
Prohibited uses are broadly drawn. The terms forbid illegal activity including copyright infringement, malware distribution, spamming, and violating others' privacy rights, and violations may result in suspension or termination without refund. Note the interaction: termination for a terms violation carries no refund, and the vendor determines what constitutes a violation.
Traffic logging is undisclosed. The privacy policy covers registration email addresses, support correspondence and marketing preferences, and cites GDPR for EU users and CCPA for California residents with a Do Not Sell or Share option. It does not state whether proxy traffic itself is logged, what is retained, or for how long. For a service whose entire function is routing your traffic, that omission is material — and it sits alongside the AML policy's commitment to retain customer due-diligence data and transaction logs for at least five years.
Aproxy competes in a crowded market against providers including Oxylabs, Decodo, Bright Data and a long tail of smaller resellers. Its positioning is on price and breadth rather than measured performance, which is a rational strategy for a challenger but has consequences for how you should evaluate it.
The established providers in this category have submitted to independent benchmarking, and their success-rate and response-time figures appear in published comparative tests. Aproxy has not. If your selection process weighs verified performance, that gap is decisive on its own. If your process weighs entry price and payment flexibility — particularly the cryptocurrency and Alipay options, which several larger competitors do not offer — Aproxy's proposition is more competitive.
The unlimited residential tier is a genuine differentiator. Flat-rate time-based residential bandwidth is not universally available, and for the specific workload profile of high, unpredictable volume without strict geographic targeting, it can be substantially cheaper than metered alternatives. That is a real reason to consider this provider, provided the billing-unit ambiguity noted above is resolved in writing first.
If you are assessing this service, the following order will surface problems while they are still cheap to discover.
Start with the smallest purchasable quantity, never a large commitment. Run your actual target sites, not a generic connectivity test — success rates vary enormously by target and an aggregate uptime figure tells you nothing about your specific case. Do all of this inside the first 48 hours, because that window is where refund eligibility lives. Confirm the billing unit of any time-based plan in writing with support before purchasing. If you have a compliance obligation, request the IP sourcing chain in writing and treat a non-answer as an answer. Verify exit geography with the proxy checker rather than trusting the country label. And keep your own logs of failures with timestamps, because a refund request under this policy requires you to document the problem.
The evidence supports it being a real, independently operated business. It discloses an operating entity, Smart Innovation Technology Limited, in its AML policy; that Hong Kong jurisdiction is corroborated by a Trustpilot profile claimed by the company in March 2025 listing Hong Kong as its location. All legal pages — terms, privacy, refund, cookie and AML — exist with substantive content, and the site runs 682 English URLs plus seven other language trees. No brand owner has named it as an impersonator and no security research has linked it to a malicious network. That is a materially different profile from a shell site with missing legal pages.
Aproxy states the pool is ethically sourced from carefully selected partners upholding business ethics and strict compliance, and that a majority consists of what it calls Tier A model proxies. The partners are not named, "Tier A model" is not a defined industry term, and the consent mechanism by which end users agree to have their bandwidth resold is not described. The claim is unverified rather than contradicted. If provenance matters to your compliance position, request the supplier chain in writing before buying.
Entry prices are $0.50/GB for rotating residential, $67.87/day for unlimited residential, $0.27/GB for long-acting ISP, $0.27/IP for static residential, $0.18/IP for static datacenter, $0.39 per 1,000 results for the Web Scraper API, $0.75 per 1,000 for the SERP API, and $0.06/GB for the Video Data API. These are "starting at" rates tied to larger commitments; your actual rate at low volume will be higher and requires an account to see.
Because the site is internally inconsistent. The residential page shows both $0.50/GB and $0.8/GB, and the unlimited page renders its rate as both $67.87/Day and $67.87/Hour. The second discrepancy is a factor of twenty-four and one of the two figures must be wrong. The public site does not resolve which. Confirm the billing unit with support in writing before purchasing a time-based plan.
Sometimes, within tight limits. Refunds are granted for total connection failure or unusable IPs unresolved within 48 hours, systemic outages exceeding 24 consecutive hours, and duplicate payments. They are refused for configuration errors, conflicts with your own network or software, orders with significant traffic consumed or used beyond 48 hours, and free or promotional packages. The company holds final interpretation. Test immediately after purchase — the 48-hour window is the real constraint.
Match the type to the session requirement. Rotating residential suits high-volume collection where each request can come from a different address. Long-acting ISP, with sessions up to 12 hours, suits anything requiring a persistent logged-in state. Static residential suits work needing the same address presented consistently. Static datacenter is cheapest and fastest but easiest for target sites to detect. Unlimited residential suits high, unpredictable volume where random country assignment is acceptable.
Through standard HTTP or SOCKS5 proxy endpoints, authenticated by username and password or by IP whitelist. Anything accepting a proxy configuration will work. The help centre documents operating system and browser setup, and covers integration with the BitBrowser anti-detect browser and V2Ray. A Public API allows programmatic proxy management, and a dashboard handles sub-users and whitelists.
No. Both are vendor self-reported. Proxyway, the most established independent proxy benchmarking outlet, does not include Aproxy in its tested rankings, and no other independent audit of these figures was found. Trustpilot carries only two reviews total, which supports no conclusion. Treat these numbers as marketing claims and verify performance yourself against your own targets before committing.
Yes. The FAQ states that for policy reasons the proxy must be used in a non-mainland-China internet environment. The terms prohibit illegal use including copyright infringement, malware distribution, spamming and violating others' privacy rights, and violations may bring suspension or termination without refund. The AML policy contemplates identity verification where red flags such as unverifiable identification or refusal to provide business details appear.
The published privacy policy does not say. It addresses registration data, support correspondence and marketing preferences, and cites GDPR and CCPA rights, but is silent on whether proxy traffic is logged and for how long it is retained. Separately, the AML policy commits to retaining customer due-diligence data and transaction logs for at least five years. If traffic logging is material to your use, ask support directly and get the answer in writing.
