Send a URL. Get the page.Even when your own code can't.
Some pages arrive empty, because JavaScript builds the content after the page loads. Others answer 403, because the request came from a server and not a browser. Send those to us and you get the page: markdown, HTML, text, links, a screenshot or a PDF. We run the browsers and the proxies.
Free tier, no card. It refills every month.
A real scrape, on our account. The same call the command below makes.
Press Run. The page shows up right here.
curl "https://api.scrapa.app/v1/scrape?url=https%3A%2F%2Fexample.com" \
-H "Authorization: Bearer sc_your_key"One key. Four things it does.
Most people buy these from four different companies. Here it's one signup, one balance, one bill.
- Grab a page/v1/scrape
- Send a URL. Get markdown, HTML, text, links, a screenshot or a PDF.
- Search the web/v1/search
- Ask a question, get up to 10 results with the dates they were really published.
- Drive a browser/v1/sessions
- A real browser on our machine that you control with Playwright or Puppeteer.
- Come from anywhere
- Pick the country your requests go out from, down to the city. Or plug in the proxy you already pay for.
It tries the cheap way first.
engine: "auto" starts with a plain request. Some pages come back empty, with nothing but a note asking you to turn JavaScript on. Those we load again in a real browser. You pay for whichever one actually worked.
- A plain HTTP request1cr
- A real browser page load5cr
You never pay for both. If the browser can't get it either, the credits go back. You keep whatever the first try found.
Need to click around? Take the whole browser.
Some jobs don't fit in one request: a login, a form, three pages deep. Ask for a session and you get a real browser running on our machine. You drive it with Playwright or Puppeteer. It is the same code you'd write on your laptop. We just hand you the address to connect to. You pay by the minute, only while it's open.
Nothing new to learn and no wrapper to fight. It's the browser you already know, running somewhere else.
Ask the web a question.
Send a question to /v1/search. You get up to 10 results, each with the date it was really published. Ask from Brazil and you get what Brazil sees, in Portuguese. We run the search ourselves. That is why we can give you the results at all.
Want it summed up? Ask, and we'll write a paragraph over the results. Ask the same question twice within ten minutes and the second one is free.
Bring your own proxy. Free, on every plan.
By default your requests go out through home internet addresses. A site sees an ordinary visitor, not a server. You pick the country, or a state or city inside it, and the first megabyte of every request is included. Already paying for a proxy? Give us the URL and we'll use yours instead. It works on the free plan too, and we never charge for its traffic.
You already pay your provider. Charging you a second time for the same bytes is not a business model. It is a toll.
When it doesn't work, you find out why.
Every answer names who said no: the site, our rules, or us. Three of the four cost you nothing, unless your own script ran in the page first: then only our own mistakes stay free.
- The page answered1cr
- You get the page. It is the only one you pay for, unless your own script ran in the page first.
- The site said no0cr· blocked · not charged
- The site put up a wall, or asked for a captcha. Try again with a real browser, from another country, or through your own proxy.
- Our policy said no0cr· denied · not charged
- The address points somewhere we won't go. It might be a private network, or a site on our no-go list. Change the URL; retrying won't help.
- We broke it0cr· refunded
- Something broke on our side. Your credits come back on their own. This is the one worth running again exactly as it was.
Charging you for a page you never got is charging you for a failure. Every company in this business says they don't. We put the number in the response, so you can check instead of trusting us.
Your AI agent can use it too.
Everything here is also an MCP endpoint. Point Claude, Cursor or your own agent at it with the same key. It runs inside the same quota and the same limits. The docs come as llms.txt too, because these days the reader is often not a person.
{
"mcpServers": {
"scrapa": {
"type": "http",
"url": "https://api.scrapa.app/mcp",
"headers": { "Authorization": "Bearer sc_your_key" }
}
}
}- One endpoint, streamable HTTP, no session to manage.
- Every operation and every refusal, machine-readable.
We read the page in São Paulo.
The browsers and the database run in Brazil. A screenshot or a PDF comes back inside the response and is never kept. We keep a page we read for you only when you ask (with maxAge or a batch), and only for your account.
9 rows. Know the price before you send a request.
One file the API prices from, published here. Any request can be priced without making it. Every answer says what it cost.
| Unit | Credits |
|---|---|
| A page read with fetch | 1 |
| Every URL a site lists | 2 |
| A web search, up to 10 results | 4 |
| A paragraph written over a search's results | 1 |
| A page rendered in the browser* | 5 |
| The fields you asked for, read off the page | 5 |
| One minute of a live browser session | 2 |
| Bandwidth through the managed pool, per MB** | 10 |
| Bandwidth through your own proxy | 0· never metered |
- *
- Screenshots and PDFs included, at no extra charge.
- **
- The first 1 MB of every request is included.
What costs nothing
A page the site blocked. A URL our policy refused. Anything we broke. All three answer with a cost of 0, and you can check it on the response itself. If your own script ran in the page first, only the last one stays free.
Plans
Credits arrive at the start of every billing month and do not roll over. Purchased credits are a separate balance that never expires.
Free
$0/month
- 1,000 credits a month
- 2 scrapes at once
- 1 live session
- Sessions up to 15 minutes
- 500 MB through the managed pool
- 25 URLs per batch
- 2 API keys
- 60 requests a minute
Starter
$19/month
- 50,000 credits a month
- 5 scrapes at once
- 3 live sessions
- Sessions up to 60 minutes
- 3,000 MB through the managed pool
- 500 URLs per batch
- 5 API keys
- 300 requests a minute
Pro
$49/month
- 150,000 credits a month
- 10 scrapes at once
- 6 live sessions
- Sessions up to 120 minutes
- 10,000 MB through the managed pool
- 2,000 URLs per batch
- 10 API keys
- 600 requests a minute
Scale
$99/month
- 400,000 credits a month
- 20 scrapes at once
- 10 live sessions
- Sessions up to 240 minutes
- 25,000 MB through the managed pool
- 10,000 URLs per batch
- 25 API keys
- 1,200 requests a minute
First page in five minutes.
Create an account. Mint a key. Paste the curl. No card, and the free tier refills every month.
Get a key