Rotating residential proxy API for web scraping

How does the residential proxy API work?

You send every request to one gateway, and the gateway forwards it through a residential IP that matches the username you sent. There is no list of IPs to download and no rotation code to write.

1. Authenticate with the username and password of your plan or of a sub-user, or from an allowlisted IP.
2. Choose the exit by adding parameters to the username: -country-de, -city-berlin, or for the United States -state-ny and -zip-10001; -asn-7922 picks an internet provider at no extra cost. City availability varies.
3. Choose the session: without a session ID, each new connection gets an IP from the pool, usually a different one; -session-<id>-ttl-<minutes> keeps one IP for up to 30 minutes, though the IP can change earlier.

The same username works on the HTTP port (7777) and the SOCKS5 port (7778). HTTPS sites go through a CONNECT tunnel, so TLS runs between your client and the website.

import requests

GATEWAY = "gw.crawlrelay.com:7777"


def proxy(username, password):
    url = "http://%s:%s@%s" % (username, password, GATEWAY)
    return {"http": url, "https": url}


# Rotating: each new connection usually exits through a different residential IP in Germany
r = requests.get("https://api.ipify.org", proxies=proxy("USERNAME-country-de", "PASSWORD"), timeout=(10, 30))
print("rotating exit:", r.text)

# Sticky: the same IP for up to 30 minutes, across connections (it can change earlier)
sticky = proxy("USERNAME-country-de-city-berlin-session-crawl42-ttl-30", "PASSWORD")
for _ in range(3):
    print("sticky exit:", requests.get("https://api.ipify.org", proxies=sticky, timeout=(10, 30)).text)

Rotating or sticky: which session should a scraper use?

Rotate for independent pages; stay sticky for flows. A crawler fetching thousands of unrelated product or article pages should open a new connection per page, without a session ID, so the load spreads over many residential IPs. A flow whose next step depends on the previous one, such as pagination with a cursor tied to a cookie, should keep one sticky session until the flow ends.

Rotation happens per new connection, not per HTTP request. A client that keeps one connection alive keeps the same exit for every request on it. The sessions guide shows how to control this in requests, httpx, Scrapy and Playwright.

What success rate should I expect from residential proxies?

Expect a high connection rate and a lower success rate on protected targets, and measure both. Proxyway's Proxy Market Research 2026 tested 13 providers: the median residential network answered 99.28% of requests, but only 74.43% of requests to real targets such as Amazon, Google and Instagram succeeded.

The gap is the target's own filtering, which no proxy removes. We publish no success-rate figure of our own until we have measured one with a stated method. Compare providers on cost per successful request for your own targets: run the same 1,000 URLs through each, count usable responses, and divide the traffic cost by that count.

Where do the residential IPs come from?

From established partner networks. CrawlRelay does not run its own device network: it resells IPs from established upstream partner networks. Our upstream provider confirms that device owners opt in and can opt out, and that its supply does not come from seized or disrupted networks such as NetNut or IPIDEA. That is the provider's confirmation, not a check we run on every device; the sourcing page lists its limits. The 2026 takedowns showed what is at stake: in January Google took legal action against IPIDEA domains, and in July the FBI and IRS-CI seized NetNut's domains, after Google reported that devices had joined these networks without their owners' knowledge. Ask any provider, including us, where its IPs come from before you build a pipeline on it.