ScrapingAPI Browser Instructions

Render a page in a real browser, tell it what to wait for or do, and get the page plus any captured network response back as JSON.

Endpoint

POST https://scrape.proxyrack.net/api/v1-beta/?api_key=YOUR_KEY Content-Type: application/json

The API key goes in the query string, everything else in the JSON body. The existing GET /api/v1/ endpoint is unchanged.

Request

{   "url": "https://example.com/page",   "country_code": "us",   "device_type": "desktop",   "premium": true,   "referer": "https://www.google.com/",    "render": {     "instructions": [ ... ]   } }

Send render to load the page in a browser, or http for a plain request. One or the other, not both.

Field	Meaning
`country_code`	Optional exit country, ISO 3166-1 alpha-2, lowercase.
`device_type`	`desktop` (default) or `mobile`.
`premium`	Default `true`.
`referer`	Sent as the navigation referrer.
`render.instructions`	Steps the browser runs after the page has loaded, in order. `"render": {}` renders with no steps.
`http.method`, `http.body`	Plain mode only: request method and body, for form or JSON posts.

Instructions

Every step is {"type": ..., ...}. Times are seconds, fractions allowed. Each wait needs a timeout; the whole plan must fit in 60 seconds.

Type	Fields	Does
`wait`	`seconds`	Pause.
`wait_for_selector`	`selector`, `state`, `timeout`	Wait until an element matching `selector` is `attached` (default), `visible`, or `hidden`.
`wait_for_response`	`url_contains`, `method`, `timeout`, `capture`	Wait for an XHR or fetch whose URL contains the text. With `capture: true` its status, headers and body are returned in the step result.
`wait_for_navigation`	`timeout`	Wait for the page to navigate to a new document, e.g. after a click.
`wait_for_load`	`state`, `timeout`	Wait until the current document reaches `load` (default), `domcontentloaded` or `networkidle`.
`click`	`selector`	Scroll the element into view and click it.
`input`	`selector`, `value`	Focus the element and type the value.
`evaluate`	`script`, `timeout`	Run JavaScript in the page; the value you `return` comes back as the step result. Async and `await` work.

Selectors are one of {"css": "#id"}, {"xpath": "//div[@id='id']"}, or {"text": "Track shipment"} (substring of the rendered text).

may_fail: true on any step lets the plan continue if that step fails; the error is still reported on the step.

Response

{   "status_code": 200,   "resolved_url": "https://example.com/page",   "headers": [{"name": "content-type", "value": "text/html"}, ...],   "body": "<html>...",   "render": {     "steps": [       {"type": "wait_for_response", "duration": 4.7,        "result": {"url": "...", "status_code": 200, "headers": [...], "body": "{...}"}}     ],     "xhr": [{"url": "...", "method": "POST", "status_code": 200, "size": 4381}, ...]   } }

status_code, resolved_url, headers and body describe the final document, after redirects and navigations. body is the rendered DOM.
steps has one entry per instruction that ran, in order, with its duration and, for wait_for_response with capture and for evaluate, a result.
xhr lists every XHR and fetch the page made, so you can find the data call to capture.

Example: MSC container tracking

The page loads its result through a TrackingInfo request after load. Capture that instead of parsing the DOM:

curl -s 'https://scrape.proxyrack.net/api/v1-beta/?api_key=YOUR_KEY' \   -H 'Content-Type: application/json' \   -d '{     "url": "https://www.msc.com/en/track-a-shipment?params=dHJhY2tpbmdOdW1iZXI9RUJLRzE4MDM5MjgyJnRyYWNraW5nTW9kZT0x",     "premium": true,     "render": {       "instructions": [         {"type": "wait_for_response", "url_contains": "TrackingInfo", "timeout": 40, "capture": true}       ]     }   }'

The carrier’s JSON is in render.steps[0].result.body. Typical time: about 10 seconds. params is base64 of trackingNumber=EBKG18039282&trackingMode=1; both keys must be present or the page does not search.

Notes

One render is one browser on one exit IP; the page and its XHRs share both.
Custom request headers and cookies are not accepted in this beta; referer is the one request header you can set.
Unknown fields, unknown instruction types and legacy instruction shapes are rejected with 400 naming the field.