Web Search
Google, Bing, and DuckDuckGo merged, deduplicated, and re-ranked into one scored result set. Pin one engine or use all three.
POST /v1/search
640ms
cold · 80 concurrent
live · serving requests
Search, scrape, RAG — behind one key.
One bearer token for multi-engine web search, any-URL scraping, RAG answers with citations, Google Maps reviews, and audio transcripts. Anti-bot walls and IP blocks are someone else's problem now. Failures don't bill.
- ~640ms cold search, P50
- ~100ms cached, P50
- 1,000 free credits, no card
P50 latencies; your numbers will vary by region and target.
the catalog
Hard targets. Already solved.
Every endpoint returns a stable schema. The proxy tier, JS rendering, cache TTL, and anti-bot handling are picked for each source. You just see clean JSON.
Deep Search
Search + auto-fetch the top N pages with extracted text. One round-trip for RAG.
POST /v1/search/deep
1 round-trip
AI Answer
RAG-style. Search, fetch, summarise with numbered [1][2] citations.
POST /v1/search/answer
real cited sources
Any URL
Point us at a page. We get past the wall (anti-bot, IP blocks, JS-heavy pages) and hand back the rendered HTML and clean text.
POST /v1/scrape/url
anti-bot · auto-tier
Translate
28 languages. Flat per-call price, no per-character or per-word billing. A drop-in alternative to DeepL or Google Translate.
POST /v1/translate
28 languages
Google Maps reviews
Up to ~10,000 reviews per place as structured JSON, newest first. 100 credits per 50 reviews returned (40 if 50 or fewer). Author, rating, text, posted_at, photos, owner reply.
POST /v1/scrape/google-maps/reviews
up to ~10k / place
Audio transcripts
Timestamped chunks for any audio URL — YouTube, X Spaces, SoundCloud, Apple Podcasts, direct MP3. Sync and async variants. 20 credits per 2 minutes of audio.
POST /v1/audio/transcript
async · no timeout
Browse the full catalog
Batch URLs, screenshots, PDF, and GasBuddy prices — plus everything else, one stable schema.
See all 11 endpoints →Catalog grows. Vote on the next endpoint or pull anything else through the generic URL fetcher.
anatomy of a request
Bearer token in. Clean JSON out.
You write the curl. We do the unglamorous half (anti-bot challenges, IP blocks, parsing five flavours of broken HTML) and hand back stable, versioned JSON.
-
01
You send a POST
Auth verified in ~50 ms. Rate-limited per key, per minute, per day. Failed requests don't bill.
-
02
We pick the tier
Datacenter for soft targets, residential for hostile ones, a real headless browser for challenge walls. We choose per-domain by default, or you can pin the tier yourself per request.
-
03
Cache or fetch
Same request inside the TTL? Cached hit at 100 ms for 1 credit. Otherwise we fetch, extract, normalise, and cache it for next time.
-
04
JSON, every time
Stable response schema per endpoint. Pre-built scrapers don't make you write selectors. Failures return credits.
curl -X POST https://scrapenest.dev/v1/search \ -H "Authorization: Bearer sn_••••" \ -H "Content-Type: application/json" \ -d '{ "query": "best wireless earbuds 2026", "num_results": 5 }'
{
"query": "best wireless earbuds 2026",
"results": [
{
"position": 1,
"url": "https://www.rtings.com/headphones/reviews/best/wireless-earbuds",
"title": "The 6 Best Wireless Earbuds - Spring 2026",
"snippet": "Our top pick is the Sony WF-1000XM5...",
"engine": "bing",
"engines": ["bing", "google"],
"score": 0.98
},
/* 4 more */
],
"took_ms": 640,
"cache_hit": false,
"credits_charged": 5
}
pricing
Pick a bundle. Or pay per credit.
Every plan has every endpoint, every proxy tier, JS rendering. The higher the bundle, the cheaper each credit. Failures don’t bill.
Free
Kick the tires
$0forever
- 1,000 credits / mo
- 1 concurrent
- All endpoints
Starter
Side projects
$29/ mo
- 200,000 credits / mo
- 10 concurrent
- Geotargeting
Growth
For shipping
$99/ mo
- 1,500,000 credits / mo
- 40 concurrent
- Priority queue
Pro
For teams
$249/ mo
- 4,000,000 credits / mo
- 100 concurrent
- Priority support
- 5 seats
- Custom endpoint requests
Scale
Data teams
$599/ mo
- 10,000,000 credits / mo
- 200 concurrent
- SLA + on-call
- Unlimited seats
- Custom endpoint requests
Or pay as you go
No commit. $0.0002 per credit, top up as you go. Same endpoints, same proxy tiers.
Plain HTTP
Datacenter, no JS
1 credit
$0.0002
JS rendering
Headless browser
5 credits
$0.0010
Residential proxy
Geo-targeted, no JS
10 credits
$0.0020
Residential + JS
Hostile targets, full render
40 credits
$0.0080
Cache hit
Repeated inside TTL window
1 credit
$0.0002
Need 25M+ credits or a dedicated VPC? Talk to sales. Plans cancel anytime, no contracts.
questions
Answered.
How is this different from a generic scraping API?
Generic scraping APIs hand you a URL fetcher: you bring the selectors and write the parsers. ScrapeNest centres on pre-built endpoints for specific high-value sources (Google Maps reviews, audio transcripts). Stable schema, no selectors, credit cost matches what the call actually used. The generic scrape/url covers the long tail.
What happens when a target hard-blocks the request?
If our proxy chain is exhausted and the source still refuses (CAPTCHA wall, geofence, hard-block), the request returns 502 and the credits return to your balance inside 60 seconds. Failures don't bill.
Do I need to bring my own proxies?
No. Proxy bandwidth is bundled. We route each call through the tier that fits the source automatically: datacenter for soft targets, residential for hostile ones, headless browser when there's a challenge page. Override per-request with proxy_tier.
Can I use this for ML training data?
Yes. That's a primary use case. Growth and Scale customers run daily ingest jobs into RAG indexes and fine-tuning corpora. Talk to us about volume rates above 500k req/mo.
ready when you are
Ship your scraper today. Sleep tonight.
1,000 free credits. No card. The API key prints once. Copy it, paste it, you're done.