Pick the right proxy. Save 5× on credits.

A proxy sits between your scraper and the target. Every request is forwarded through it; the target sees the proxy’s IP, not yours. That gives you three things: anonymity (your real IP isn’t on the target’s logs), distribution (no per-IP rate limits when you rotate), and geo-targeting (request from the country the site expects).

Sites monitor request patterns per IP. Too many requests from one IP → rate limit, soft block, CAPTCHA, or hard ban. Different proxy types exist because different sites do different kinds of monitoring. The 80/20 isn’t about which proxy is “best” — it’s about matching tier to target.

Skip the rest of this page if you only read one section. Match your target to the row that fits — the recommended stack is in the middle column. Real-world success rates measured across our customers in Q1 2026.

The trade-offs are linear in obvious ways and non-linear in subtle ones. Datacenter is 5× cheaper than residential but only ~5% lower in success rate on most sites — making it almost always the right first attempt. Mobile costs 10× datacenter and only matters on a small set of targets where the marginal trust unlocks the page.

The same six categories show up across every provider — naming varies, behaviour doesn’t.

auto_proxy=true implements the decision tree at request time. The loop is: try cheap, escalate on failure, return on success. You’re only billed for the request that actually returned a 2xx — failed escalation steps are free.

Two questions decide which one fits: (1) does the target track sessions, and (2) does the scrape need cookie/state continuity across requests?

Per-call rotation (default)

Fresh IP per request. Best for high-volume stateless scraping — product catalogs, SERP tracking, news monitoring. No per-IP rate limits, no extra cost.

Sticky sessions

Pin an IP for up to 30 min via proxy_session="sess_42". Required for logins, multi-step forms, and paginated results that bind to IP. Same key reuses the same exit IP.

“US proxy” is rarely specific enough. Amazon shows different prices in California vs Texas. Google SERPs vary city to city. A US proxy on a US site triggers fewer challenges than a foreign IP — match the geo, and the rest of the stack relaxes.

curl -X GET 'https://api.example.com/scrape' \
  -H 'ApiKey: YOUR_API_KEY' \
  -G \
  --data-urlencode 'url=https://www.amazon.com/dp/B08N5WRWNW' \
  --data-urlencode 'premium_proxy=true' \
  --data-urlencode 'proxy_country=US'

Pool granularity by tier: datacenter ≈ country, residential ≈ city, mobile ≈ carrier-region (often 1–2 metro areas). For city-level Google SERPs, residential is the floor.

All proxy configuration is request parameters — no separate proxy SDK to wire up.

Auto-proxy with failover (the cheap default)

curl -X GET 'https://api.example.com/scrape' \
  -H 'ApiKey: YOUR_API_KEY' \
  -G \
  --data-urlencode 'url=https://example.com' \
  --data-urlencode 'auto_proxy=true'

Residential + country targeting

import requests

response = requests.get(
    'https://api.example.com/scrape',
    params={
        'url': 'https://protected-site.com',
        'premium_proxy': 'true',
        'proxy_country': 'US',
    },
    headers={'ApiKey': 'YOUR_API_KEY'},
)
print(response.text)

Sticky session for paginated scraping

// Same session key = same exit IP across the loop
for (let page = 1; page <= 5; page++) {
  await fetch('https://api.example.com/scrape?' + new URLSearchParams({
    url: `https://example.com/page/${page}`,
    premium_proxy: 'true',
    proxy_session: 'my-session-123',
  }), { headers: { ApiKey: 'YOUR_API_KEY' } });
}