Residential proxies for AI training data collection

High-volume crawling setup

Large crawls are stateless: point your crawler at the rotating port 9000 so each request uses a new IP. If your pipeline speaks SOCKS5, the same credentials work on port 11000. Use socks5h so hostnames resolve through the proxy.

# HTTP, rotating
curl -x http://<USERNAME>-type-residential-country-us:<PASSWORD>@gw.nodelighthouse.com:9000 https://example.com/article/1

# SOCKS5
curl -x socks5h://<USERNAME>-type-residential-country-us:<PASSWORD>@gw.nodelighthouse.com:11000 https://example.com/article/1

Multilingual and regional content

Many sites serve a different language or edition by location. Set the country in the username to collect the regional version directly, which helps when you need balanced data across languages.

Collect responsibly

Dataset collection has legal and ethical limits. Collect publicly accessible content, honor robots.txt and opt-out signals, and respect copyright and licensing for how the data will be used. Don't mass-copy copyrighted content for redistribution, and handle any personal data in line with applicable privacy law. See the acceptable use policy.

• Deduplicate and filter before storing to save bandwidth and storage.
• Skip binary assets you won't train on: images, video and fonts all count toward GB used.
• Keep per-domain concurrency low and back off on errors.

Cost at scale

Pricing is $1.65/GB on every plan, including 1000 GB for $1,650, so cost scales linearly with the data you fetch. Payment is in crypto. See pricing.

Setup guides

• Scrapy
• Python requests
• Go (net/http)
• Playwright

Questions

Is there a volume discount?

No. Every plan is a flat $1.65/GB, from 5 GB up to 1000 GB.

Can I export a proxy list for a distributed crawler?

Yes. The dashboard proxy generator exports up to 1000 lines in host:port:user:pass and other formats.

Which payment methods are accepted?

Crypto only, including USDT, USDC, BTC, LTC, SOL, TRX and ETH.