Multi-engine web search, any-URL scraping past anti-bot walls, cited AI answers, Maps reviews, and audio transcripts — one bearer token, clean JSON. MCP-native for Claude, Cursor, and ChatGPT. Failures don't bill.
pip install scrapenest
npm install scrapenest
MCP · Claude · Cursor · VS Code · ChatGPT
the catalog
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.
Google, Bing, and DuckDuckGo merged, deduplicated, and re-ranked into one scored result set. Pin one engine or use all three.
POST /v1/search
Search + auto-fetch the top N pages with extracted text. One round-trip for RAG.
POST /v1/search/deep
1 round-trip
RAG-style. Search, fetch, summarise with numbered [1][2] citations.
POST /v1/search/answer
real cited sources
Connect Claude, Cursor, VS Code or ChatGPT. Web search, scraping and translation become native tools. One endpoint, your key. Tools bill the same as the API.
POST /mcp
Claude · Cursor · ChatGPT
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
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
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
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
Batch URLs, screenshots, PDF, and gas 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.
mcp-native
One URL, your key. Claude, Cursor, VS Code, and ChatGPT gain live web search, scraping, deep research, and translation as built-in tools. Billing is identical to the API.
claude mcp add scrapenest --transport http \ --header "Authorization: Bearer YOUR_KEY" \ https://scrapenest.dev/mcp
{
"mcpServers": {
"scrapenest": {
"url": "https://scrapenest.dev/mcp",
"headers": { "Authorization": "Bearer YOUR_KEY" }
}
}
}
{
"servers": {
"scrapenest": {
"type": "http",
"url": "https://scrapenest.dev/mcp",
"headers": { "Authorization": "Bearer YOUR_KEY" }
}
}
}
https://scrapenest.dev/mcp?profile=chatgpt&apiKey=YOUR_KEY
Add as a connector in ChatGPT (Developer Mode). The chatgpt profile exposes the search + fetch tools deep research requires.
web_searchmulti-engine, merged & rankedask_webcited answer from live pagesdeep_searchsearch + full text of top resultsscrape_urlJS + anti-bot handledtranslate_text28 languages, flat priceanatomy of a request
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.
Auth verified in ~50 ms. Rate-limited per key, per minute, per day. Failed requests don't bill.
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.
Same request inside the TTL? Cached hit at 100 ms for 1 credit. Otherwise we fetch, extract, normalise, and cache it for next time.
Stable response schema per endpoint. Pre-built scrapers don't make you write selectors. Failures don't bill.
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": 1310,
"cache_hit": false,
"credits_charged": 5
}
pricing
Every plan has every endpoint, every proxy tier, JS rendering. The higher the bundle, the cheaper each credit. Failures don’t bill.
Kick the tires
Side projects
For shipping
For teams
Data teams
No commit. $0.0002 per credit, top up as you go. Same endpoints, same proxy tiers.
Need 25M+ credits or a dedicated VPC? Talk to sales. Plans cancel anytime, no contracts.
questions
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.
If our proxy chain is exhausted and the source still refuses (CAPTCHA wall, geofence, hard-block), the request returns 502 and no credits are charged. Undeliverable requests simply don't bill.
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 render and force_proxy.
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.
1,000 free credits. No card. The API key prints once. Copy it, paste it, you're done.