Oxylabs Proxy Review: Pricing, Networks, Scraping APIs and Tradeoffs

Oxylabs verdict: broad capabilities with product-specific costs

Oxylabs is most compelling when a project needs several forms of access, precise geography or a choice between owning a crawler and delegating retrieval. A team can begin with a proxy-connected client, move difficult pages to managed retrieval, and use a remote browser for interactive tasks. Each move changes what the provider operates and how usage is billed, so the product portfolio should be treated as separate implementation options.

Oxylabs — proxy networks and managed web data tools

Choose the service around the job

Use this homepage image as a visual introduction; screenshots may retain older claims. Current package details appear below. Start with the network, session requirements and output format you actually need.

ResidentialMobileDatacenter / ISPManaged retrievalBrowser automation

Explore Oxylabs products

For a wider shortlist, use our best proxy providers guide. Compare providers on the same targets and acceptance criteria, rather than treating a larger advertised IP count as proof of better data or faster requests.

Proxy networks and which product to choose

Rotating residential and mobile

Residential advertises 175M+ IPs across 195 countries. Mobile advertises 20M+ IPs across 140+ countries and 3G/4G/5G/LTE routes. These are vendor pool descriptions, not our count of live exits. They do not establish how many peers are simultaneously available in a small city or carrier network. A narrow geographic test is more useful than assuming every location has the same depth. Compare residential versus datacenter proxies when weighing IP origin, sourcing, consistency and cost.

Choose residential when geographic routing and consumer-network assignment match the task. Choose mobile when cellular-network behavior is a real requirement, such as mobile ad verification or authorized application testing. Paying a mobile premium without testing a mobile-specific benefit can increase cost without improving accepted records. See our mobile proxy server guide for the wider selection criteria. Use our guide to proxy server types to separate IP source, protocol and session behavior before comparing plans.

Datacenter and ISP

Datacenter separates dedicated per-IP, shared per-IP and shared rotating per-GB access. The public cards list 188, 38 and 44 countries respectively, so the broadest catalog coverage should not be assigned to the cheapest shared tier. Those cards also describe different pool sizes; there is no reason to add them into a performance score or assume the same allocation guarantees.

ISP addresses are registered under residential ISPs while hosted on server infrastructure. Shared ISP is shared with up to three users, while dedicated ISP is allocated to one customer. Both list 25 locations and unlimited-duration sessions under their product terms. This is useful for continuity, but does not make a rotating residential session permanent or remove a target’s ability to classify and restrict traffic. Our proxy rotation and session guide explains how rotating exits differ from keeping a sticky session.

Managed retrieval, parsed output and interactive browsing

Web Unblocker suits an existing application that wants managed page access while retaining its own parser. Web Scraper API can return raw content or structured data for supported sources, with custom parsing and collection features. Headless Browser suits workflows that must click, scroll or maintain browser state. Its current pricing page uses Headless Browser, while some navigation uses Agent Browser; confirm the selected service and endpoint instead of assuming a different pricing model from a label.

Our web scraping tools guide helps compare the operational scope. A proxy, a page retrieval service and an end-to-end data platform solve different portions of a pipeline even when all can eventually contribute to the same dataset.

Current Oxylabs proxy pricing

The tables show public monthly package totals before any applicable VAT. Published unit labels can be rounded: Mobile Basic is $100/15 GB, approximately $6.67/GB, and Corporate is $2,500/715 GB, approximately $3.50/GB. Treat the package total and included allowance as the purchase commitment; enterprise quotes, selected configurations and checkout terms may differ.

A low per-GB rate is useful only when the capacity and service match your demand. Corporate residential’s $2.50/GB headline requires a $2,500 monthly package. For a 5 GB experiment, Starter’s $30 total is the relevant commitment. Top-ups are advertised, with limits varying by tier, but confirm the incremental price, usable period and renewal behavior in the account before forecasting extra traffic.

The per-IP datacenter cards advertise unlimited bandwidth subject to fair use; the rotating package has a specified GB allowance. This is why “datacenter from $0.59” is incomplete without a unit. Neither model includes an assurance that a protected target will accept the route. Confirm geography, allocation, replacement and throughput requirements for the chosen selector.

At 100 IPs, dedicated ISP costs $140 more monthly than shared ISP. That is a 107.7% premium over the shared total for exclusivity and different fair-use terms, rather than a measured speed improvement. If the task does not need exclusive allocated addresses, test whether the shared product meets your acceptance and session requirements first.

Web Unblocker, Web Scraper API and browser pricing

Managed tools can reduce the work of operating access infrastructure, but their billing is different from raw proxies. Compare the cost of a usable page or accepted record after accounting for rendering, media and validation. A price in dollars per thousand results cannot be placed on the same numerical axis as dollars per GB or dollars per IP.

Web Unblocker measures both request and response traffic for successful extraction; provider-system failures with 5xx statuses are excluded under the billing guidance. The pricing page’s rounded list labels are $9.40, $8.60 and $7.50/GB, which do not multiply exactly to every package total. A WU40 promotion was visible during research, but its duration and eligible renewal terms were not verified. The regular totals above provide a clearer planning baseline.

Micro advertises up to 98,000 results, a headline that fits Amazon without rendering. Its public card also lists media downloads at $3/GB and a 50 requests/second rate limit. Rendering and feature-specific consumption can reduce available output. “All features included” should therefore be read alongside usage billing, rather than interpreted as an unlimited number of browser operations.

Headless Browser supports Playwright, Puppeteer and CDP-compatible clients, with managed browser infrastructure and CAPTCHA handling. Browser traffic depends on page weight, interactions and resource loading. Measure a representative complete task rather than budgeting only the initial HTML. A Web Scraper API browser-related feature and a separate browser subscription also need their own endpoint and billing confirmation.

For a managed-service comparison, read our ScraperAPI review. Match target categories, parsing requirements and rendering assumptions before comparing nominal request prices.

Bandwidth, utilization and the cost of accepted data

Purchased bandwidth is only one part of operating cost. A raw proxy can transfer an error page, repeated requests or a response your parser cannot use. A managed service can return a billable result that fails your acceptance rules. Track the provider’s usage counter and your own accepted-record counter independently; otherwise a cheaper unit rate can conceal waste elsewhere in the pipeline.

For an illustrative raw-proxy job, 10,000 pages averaging 200 KB of transferred response content equal roughly 2 GB using decimal units. Two attempts per accepted page would make that roughly 4 GB before request traffic, browser resources and billing-meter details. At a fully utilized $6/GB rate, those scenarios allocate approximately $12 and $24 of traffic cost. They are planning examples, not observed page sizes or a guarantee of a smaller invoice than the monthly commitment.

For accepted data, use total period cost divided by records that pass validation. Include subscription spend, extra traffic, compute, storage, monitoring and maintenance. If managed retrieval saves several hours of access debugging, it may justify a higher network unit price. If a plain client already succeeds on the target, browser automation may add cost without a useful benefit. Measure both routes before scaling.

Targeting, sessions, protocols and access controls

Residential lists unlimited concurrent sessions, three proxy users and ten whitelisted IPs. That does not mean unlimited traffic, infinite local capacity or permission to send unbounded traffic to a target. Keep an application-level limit, share quotas deliberately between workers, and distinguish a proxy user from a website account or an allocated exit address. Our proxy server explainer provides the request-flow background for interpreting the product claims in this comparison.

Sticky sessions need application state

A stable route is only one ingredient in a continuous browser session. Cookies, headers, locale, user agent and navigation state must also remain consistent. Rotating an exit between dependent steps can break an authorized flow; keeping an exit while discarding cookies can do the same. Test continuity across realistic task durations and define what happens when a peer disappears or a session needs replacement.

Proxy protocols are not device-wide protection

An HTTP proxy can tunnel traffic to an HTTPS destination, while SOCKS5 clients have different routing and DNS configuration requirements. Native residential HTTP/3 uses MASQUE and needs a compatible implementation. MDN’s proxy and tunneling guide explains the distinction between forwarding and tunneling. Configure the intended application explicitly; a proxy setting does not automatically cover every process on a device. Check HTTP versus SOCKS proxies before assuming an endpoint supports the protocol or DNS behavior your application needs.

Performance claims and a useful evaluation plan

Oxylabs publishes pool, response-time and success-rate claims. They describe provider measurements under particular conditions and are not comparable to an independent full-page or accepted-record benchmark. Even the residential page currently contains different latency figures in different feature blocks. This review therefore avoids numerical quality scores, speed rankings and charts that imply we measured a result we did not test.

Use a target set containing ordinary pages, heavier pages and the locations you will actually request. Run the same inputs through the candidate products, keep retry policies comparable, and validate the same fields. A trial of unrestricted easy pages does not establish access to a protected production target. Repeat enough to observe variability without treating a brief pilot as an uptime guarantee.

For browser tasks, record cold-start and subsequent-navigation timing separately, plus the resources needed to complete the task. For geographic data, check currency, locale and content region rather than relying only on an IP-checking label. A correct route that returns the wrong regional page is not successful business output.

Developer setup, dashboard and troubleshooting

Begin by selecting the purchased product in the dashboard, creating its credentials and generating the connection details. Hostnames, ports and authentication formats vary between raw proxies and managed services. Store credentials outside source control, give separate jobs separate usage visibility where supported, and establish spending alerts before starting a distributed worker pool. For a desktop configuration, follow our Windows and Mac proxy setup guide and verify which applications actually use the configured proxy.

# Install requests; obtain the complete HTTP proxy URL from your dashboard.
import os
import requests

proxy_url = os.environ['OXYLABS_PROXY_URL']
response = requests.get(
    'https://example.com/',
    proxies={'http': proxy_url, 'https': proxy_url},
    timeout=(10, 30),
)
response.raise_for_status()
if not response.text.strip():
    raise ValueError('Empty target response')
print({'status': response.status_code, 'characters': len(response.text)})
# Validate page identity and required fields before storing a record.
# Keep TLS verification enabled. Never log the credential-bearing proxy URL.

Build a collector around validated output

The example is a connectivity starting point rather than a production crawler. Add a bounded queue, cancellation, retry backoff, deduplication and a validation schema. Use an overall task deadline in addition to connect/read timeouts. Retry transient connection failures within a limit, but avoid indefinitely resubmitting a restricted target or a parsing error that another network route cannot repair.

With Web Scraper API, select the appropriate source and rendering mode, then verify the returned schema for the target. Public features include scheduling, batch queries, cloud delivery, custom parsing and OxyCopilot. Those can shorten implementation work, but parser output still needs completeness and type checks. Generated extraction logic should be reviewed against sample pages and monitored after target changes.

An existing Scrapy or request-based pipeline may need little more than integration and operational controls. An interactive browser pipeline needs a resource policy and durable task state as well. Compare those implementation needs before concluding that the broadest available tool is the best starting point.

Trials, support, refunds, sourcing and restrictions

Public plan cards list 24/7 support and a dedicated account manager. Those labels are useful purchasing signals, but do not prove a particular response time, resolution target or contractual SLA. Ask how support handles the selected tier, what information is needed for a target-access issue, and what escalation arrangement applies to a critical production job.

Refunds are conditional and differ by product

The current policy generally requires eligible self-service customers to request within three calendar days of the first transaction, with no suspension for a policy breach, chargeback or previous refund claim. Residential, mobile, dedicated ISP, Web Unblocker and Web Scraper API sections additionally limit used traffic to no more than 20% for regular plans or 10% for enterprise plans. Several per-IP sections list conditions without that traffic cap; apply the exact product section rather than extending the summary to every purchase.

Dedicated Datacenter Enterprise purchases are final and non-refundable apart from qualifying billing errors reported within 30 days. The policy says a refund may take up to 15 business days to reach the account. This is not an unrestricted money-back guarantee or a substitute for a trial. Confirm eligibility, recurring charges and cancellation terms before payment, particularly for a negotiated contract.

Fair use can change concurrent capacity

Shared ISP’s policy lists 100 concurrent sessions per IP until 50 GB per IP is reached during a billing cycle, then 10 for the remainder. Dedicated ISP states the analogous threshold as 100 GB per IP. The dedicated page’s aggregate worked examples do not consistently match that higher threshold, so confirm the governing contract and how usage is aggregated. Unlimited bandwidth should not be modeled as unlimited sustained concurrent throughput.

Sourcing and permitted targets

Oxylabs describes consent-based residential sourcing and provides supplier and trust materials for procurement review. Its pages identify Proxy Solutions and Scraper APIs as ISO/IEC 27001:2022 certified products. These are vendor disclosures; we did not audit peer consent, inspect certificate scope or perform a security assessment. Ask for the documents and contractual safeguards your organization needs, including data handling and retention details.

Restricted-target lists and KYC requirements vary by service. Residential guidance names categories including financial, government, streaming, ticketing and mailing destinations. Some access may require review, and a successful trial on one site does not imply approval for another. Establish target permission and service eligibility before buying. RFC 9309’s Robots Exclusion Protocol describes crawler rules, while explicitly distinguishing them from access authorization; robots directives alone do not settle whether a collection activity is permitted.

Use cases and recommended starting points

A mixed workload may justify several products, but route selection needs explicit rules. For example, static pages can use a plain client while a smaller interactive subset uses a browser. Record the reason for escalation and avoid silently sending every failed task to the most expensive mode. That keeps costs explainable and makes it easier to identify whether failures come from network access, page behavior or extraction logic.

For procurement, keep a target matrix with required regions, cadence, state duration, fields and delivery windows. Use our proxy comparison tool to organize provider options and the proxy deal finder to compare eligible offers after establishing the correct product and commitment.

Oxylabs pros and cons

The strongest reason to choose Oxylabs is a match between its controls or managed tools and your workload. The weakest reason is a generic claim that enterprise positioning makes every target faster or every dataset more complete. Budget a pilot that measures the benefit you are paying for.

Oxylabs alternatives and provider comparisons

A competing provider can be a better fit because of its purchase commitment, validity model, dashboard or collection scope. These differences should be evaluated alongside real target behavior. This section does not assert a provider speed ranking or use promotional prices to manufacture a universal cheapest-provider claim.

Bright Data

Alternative workflow and purchasing model

Compare the exact network or managed collection product, access checks and total operating scope. Our Bright Data vs Oxylabs comparison helps frame the choice without treating two broad catalogs as a single subscription.

Read Bright Data reviewCompare provider options

Webshare

Alternative workflow and purchasing model

Evaluate when a self-service proxy setup is enough. Compare allocation, location inventory, traffic validity and required support against the Oxylabs product you would actually purchase.

Read Webshare reviewCompare provider options

ProxyShare

Alternative workflow and purchasing model

Consider a different small-package model, while checking validity and session details. The lower commitment is useful only if the selected target and operational requirements work in practice.

Read ProxyShare reviewCompare provider options

IPRoyal

Alternative workflow and purchasing model

Investigate non-expiring residential traffic for intermittent projects. Compare unused-capacity economics, session controls and accepted output before choosing solely on the smallest-order per-GB price.

Read IPRoyal reviewCompare provider options

Homepage images are illustrative and may show earlier promotional copy. Use the detailed review and a current product quote for purchase decisions. A fair comparison keeps geography, concurrency, requested output and acceptance criteria consistent, then adds contract, support and maintenance costs.

Frequently asked questions

Research methodology and limitations

Affiliate disclosure: we may earn a commission when you subscribe through the Oxylabs links in this article. The recommendations reflect the documented product scope and workload tradeoffs explained above.

Final recommendation

Oxylabs deserves a shortlist place when precise proxy controls, allocated ISP routes or managed retrieval and browser tools can solve a real operational requirement. Its range gives teams several ways to build a collection pipeline, but choosing the right product matters more than buying into the overall catalog. A small request-based job and a large interactive browser workflow should not begin with the same assumptions.

Select the network and output requirements first, confirm the unclear billing or fair-use details, then run a representative pilot. Scale only after you understand accepted-record cost, resource consumption and maintenance effort. If intermittent usage or a very small commitment is the main requirement, compare other purchasing models before subscribing.