Skip to content
Built for AI/ML teams REST API · sticky sessions

AI Data Collection Infrastructure for LLM & RAG Pipelines

Training corpora and retrieval pipelines do not browse like humans. They fan out across hundreds of domains, retry aggressively and hold sessions for hours. ProxyPerk gives you rotating residential, mobile, named-ISP and datacenter egress behind one endpoint and one REST API — so you route per target instead of per vendor.

collect.py Python
import requests

# session string from your dashboard
SESSION = "<paste BASE64_CONFIG>"
GATE = "gate.proxyperk.com:8080"
AUTH = "user:" + SESSION
proxy = f"http://{AUTH}@{GATE}"
proxies = {"https": proxy}

r = requests.get(
    "https://httpbin.org/ip",
    proxies=proxies,
    timeout=30
)
print(r.json())
country + sticky session inside SESSION one endpoint
200 countries
35 mobile countries
23 named ISPs
99.99% uptime SLA
7-day instant cash back Unlimited bandwidth on static & VPN Live 24h abuse monitoring

Works with your stack

Python Node.js JavaScript TypeScript Go PHP Ruby Rust C++ Kotlin Swift Selenium Scrapy cURL Pytest Docker Kubernetes GitHub PostgreSQL Redis

The problem

Why Naive Collection Gets Blocked

A scraper that works from one IP fails in production for three separate reasons. Each one needs a different fix, which is why “buy proxies” is the wrong first question.

Volume limits per IP

Sites rate-limit per address. One IP at high concurrency looks like an attack regardless of how polite your crawl delay is. Rotating pools spread the same request volume across thousands of exits.

TLS and browser fingerprints

Datacenter ranges are classified by ASN before a single request lands. Residential and mobile exits carry a carrier or ISP reputation that a data-center IP cannot fake.

Geo-locked and localised content

Prices, inventory and search results differ by country, and often by city. Country targeting with city-level selection is what makes a dataset comparable across markets.

Workload routing

Route per Target, Not per Vendor

Most production pipelines use three or four classes at once. The routing decision is per domain, which is exactly what a single endpoint with per-session settings gives you.

Fan-out collection

Rotating Residential

200 countries, sticky sessions, city targeting. $3.95/GB

Held identity

Static ISP

Named carrier, never rotates. From $2.00/IP

Mobile-only sources

Rotating Mobile

35 countries, carrier NAT. $9.90/GB

Open APIs & docs

Rotating Datacenter

5 countries, fastest class. $1.90/GB

How it works

One Endpoint, Config in the Password

All three rotating types share a single gateway. Your session settings — target country, sticky session, dynamic or static password mode — are JSON-encoded, Base64-packed into the proxy password field, and unpacked server-side. You never open a second connection or ship a second credential.

1

Pick settings in the dashboard

Country, city, sticky lifetime, password mode. The dashboard emits the encoded string for that configuration.

2

Paste it as the password

The username stays the same. Only the password changes per configuration, so one client serves every session type.

3

Read usage per session

Hourly bandwidth charts split dynamic and static modes, so you can attribute cost per configuration.

Encoders ship in the dashboard for every product: Python (requests + PySocks + Base64), Node.js (socks-proxy-agent), Go (net/http + encoding/base64), PHP (cURL) and cURL. The exact JSON keys for each product are in the snippet your dashboard generates — copy it rather than hand-rolling the encoding.

Endpoint and filter reference

Developers

Drop It Into an Existing Spider

Nothing here needs a vendor SDK. Scrapy, requests, Playwright and Puppeteer all accept a standard proxy URL, so a migration or a new pipeline is a configuration change.

spider.py Scrapy
import scrapy

SESSION = "<paste BASE64_CONFIG>"
GATE = "gate.proxyperk.com:8080"
AUTH = "user:" + SESSION
PROXY = f"http://{AUTH}@{GATE}"

class CorpusSpider(scrapy.Spider):
    name = "corpus"
    start_urls = ["https://example.com"]

    def start_requests(self):
        for url in self.start_urls:
            yield scrapy.Request(
                url,
                meta={"proxy": PROXY}
            )
enable HttpProxyMiddleware in settings no SDK

Sticking a session: keep the same encoded string for the whole task and the exit IP stays put for the sticky lifetime you selected. Rotate by generating a new string — the endpoint and username never change.

At scale

What You Get Past the Prototype

The prototype is a proxy URL. Production is rotation control, usage attribution and an abuse trail you can read — all over the same API.

Rotation control

Create, list and delete rotation configurations over the API. Sessions are data, not tickets.

Usage analytics

Hourly download and upload charts per product, with dynamic and static modes tracked separately.

Abuse monitoring

A live 24h log of bans with type and reason, so you debug the target instead of the network.

Unlimited bandwidth

Static IP and VPN products are unmetered. Only the rotating pools are billed per GB.

scale.py python
import os, requests

PROXY = "http://user:BASE64@gate.proxyperk.com:8080"
os.environ["HTTPS_PROXY"] = PROXY

# one session per worker, sticky per domain
for target in targets:
    r = requests.get(target, timeout=30)
    store(target, r.text)

# usage per configuration, hourly
$ curl -H "API-KEY: $PP_KEY" \
    .../rotating_proxies/stats/
one endpoint, per-session routing 200 countries

Usage stats fields

service residential · datacenter · mobile
date range from / to
split download vs upload
mode dynamic vs static password

API reference: endpoint and filter reference

Compliance

Sourcing and Acceptable Use

If your pipeline feeds a model, your network choice is part of your compliance story. These are the published rules, not a sales line.

How IPs are sourced

  • Residential and mobile pools come from consenting device owners and licensed ISP partners.
  • Opt-in only, with a published 24-hour removal process.
  • Datasets are audited against regional regulation before entering a pool.
Ethical IP sourcing

What is not allowed

  • Fraud, phishing, credential attacks, DDoS and malware distribution.
  • Spam, and any use that breaches a target’s terms or local law.
  • Abuse is enforced with incremental bans, visible in your dashboard.
Terms of Service

Pricing

Priced per Class, Not per Bundle

Rotating pools are billed per GB. Static IP and VPN products are per IP or per server per month with unlimited bandwidth. Named ISP carriers multiply the base — the calculator shows the full breakdown before checkout.

Class Billing From
Rotating datacenter — 5 countries per GB $1.90
Rotating residential — 200 countries per GB $3.95
Rotating mobile — 35 countries per GB $9.90
HTTP/HTTPS dedicated, unlimited bandwidth per IP / month $2.00
SOCKS5 dedicated, unlimited bandwidth per IP / month $2.10
SOCKS5 shared, Dallas TX per IP / month $0.88

Enterprise anchor. Oxylabs and NetNut publish roughly $7.50–$7.60 per GB at the 50 GB tier (public pricing, November 2026). ProxyPerk sits well below that anchor, and quantity discounts start at 15 IP or 15 GB.

Open the full pricing calculator

Guarantee

Test Against Your Real Targets

A benchmark on other people’s domains tells you nothing. Buy 15 GB, run one real job against your own target list, and compare success rate and latency.

Step 1

Buy 15 GB of rotating residential or datacenter.

Step 2

Point one existing spider at the gateway — no code change.

Step 3

Compare per domain, then open a ticket if it is not as described.

Every ProxyPerk product carries a 7-day instant cash back guarantee — datacenter included, and datacenter IPs are additionally trial-eligible. The request is validated in real time against the published terms.

Covers non-delivery and service issues. Third-party software incompatibility and prohibited uses are excluded — Terms of Refund.

Help center

Questions from pipeline builders