Bytes-billed, not compute-billed

Most cloud scraping APIs charge per request, with a flat price that assumes a full chrome render. That's expensive overkill when you just want HTML. proxy-fetch uses spider.cloud's HTTP proxy network — your fetch goes through a rotating residential / ISP / mobile / datacenter IP, you get back the raw response, and you're billed for the bytes that came back plus a small per-request fee.

curl -X POST https://api.crawlcrawl.com/v1/cloud/proxy-fetch \
  -H "Authorization: Bearer crk_..." \
  -d '{
    "url": "https://target.example.com/page",
    "pool": "isp"
  }'

# → 200
{
  "url": "https://target.example.com/page",
  "status": 200,
  "content": "<!DOCTYPE html>...",
  "bytes_transferred": 184037,
  "elapsed_ms": 419,
  "cost_usd": 0.0000427,
  "pool": "isp"
}

Pool selection guide

Pricing observed against live billing 2026-05-19, accurate within ±5%:

Pool	Per request	Per GB	When to use
`datacenter`	$0.0000070	$0.05	Cheapest. Public sites with no anti-bot.
`isp` (default)	$0.0000050	$0.22	Balanced. Default for most workloads.
`residential`	$0.0000050	$0.12	Retail/social with IP filtering.
`mobile`	$0.0000050	$0.20	Sites whitelisting mobile carriers.

Per-100K-page indicative cost

Workload	Cost
100K plain HTML pages via `isp` (200 KB avg)	~$4.90
100K via `residential` with geo-routing	~$2.80
Same 100K via `/v1/cloud/scrape` chrome render	~$29.60

Geo-routing

Pass "country": "de" (ISO-2 code) to route through that country's proxy. 199 countries supported. Critical for:

Geo-localized SEO audits — see what a UK user sees vs a US user.
Localized pricing checks — most retail sites vary prices by IP region.
Compliance / regulator scrape — fetch as the regulator's expected region.

What proxy-fetch does NOT do

Honest list of trade-offs you accept for the price:

No JavaScript rendering. SPA shells return empty-ish. Use /v1/cloud/render or chrome:true on /v1/cloud/scrape for hydrated DOM.
No deep anti-bot bypass. Sites fingerprinting beyond IP (e.g. LinkedIn HTTP 999) still block. Use /v1/cloud/unblock for those.
No sticky sessions. Each call may get a different IP. Multi-step flows (login → navigate → scrape) need full chrome.

Now wired into 3 other endpoints

You can opt into proxy mode anywhere in the platform:

POST /v1/cloud/scrape with force_backend: "proxy" — single-URL cheap fetch.
POST /v1/actors/check-links with cloud_retry_mode: "proxy" — 30× cheaper retry path for broken-link audits.
POST /v1/crawls with proxy_pool: "residential" and proxy_country: "in" — route a multi-page crawl through a specific country.

Proxy Fetch.