Rotating or sticky for a crawl?

Rotating for anything where a request stands alone — sitemap walks, detail pages by ID. Sticky for anything with state: pagination, sorted search results, filters, a form and its result page. Both use the same credentials; the only difference is whether the username carries a _session- handle, and you can mix them freely in one crawl.

How long does a sticky session last?

As long as the walk needs. _sesstime- is minutes: the same _session- handle returns the same exit IP for up to that long. On ARC Basic, Fast and Ultra the hold is best effort with no hard cap — the dashboard generator offers 1 minute to 24 hours, and the API's session_ttl takes 60 to 86,400 seconds. On ARC Mobile a sticky session can be set to 5 to 60 minutes. An exit can go offline before the time is up, so treat a changed IP as normal, not as an error. Let the parser tolerate a fresh IP mid-walk.

Do you support SOCKS5?

Yes, on the same gateway and port with the same username and password — change the scheme to socks5://. Use socks5h:// in curl and requests to resolve DNS at the proxy rather than on your machine. The dashboard generator writes http:// lines; the API returns SOCKS5 lines with protocol: "socks5".

What exactly is billed?

Gigabytes of traffic through the gateway, and nothing else — not requests, connections, threads, endpoints or API calls. Generating 10,000 endpoints costs nothing; sending traffic through them costs bandwidth.

Does unused bandwidth expire?

No. Gigabytes stay on your account until you use them — no monthly reset, no rolling window, no expiry date. A quiet month costs nothing, and a bigger single purchase is only ever cheaper per GB.

Do I need a company, a call or a minimum?

No. Create an account and buy anything from 1 GB to 1 TB in one purchase; volume discounts apply automatically at checkout. A very large order is charged at checkout but can be held for a manual release before its gigabytes are credited; until then your balance shows them as pending. Above 1 TB a purchase is priced individually — write to us.

Prices on this site are in EUR and exclude VAT. VAT is calculated at checkout according to your location and status; a business can enter its name and VAT ID there, and both appear on the invoice. Every purchase is a one-time Stripe payment with an invoice from Ravaiq Ltd., and your payments and receipts are listed under Billing in the dashboard.

Can I mix plans in one crawl?

Yes. ARC Basic, Fast and Ultra share one gateway and one set of credentials; the _pool- token in the username decides which pool a connection uses. The Scrapy and Crawlee snippets above do exactly that, per domain. ARC Mobile has its own endpoints, generated for you.

Anything unlawful, plus the specific list in the acceptable use policy: fraud, account takeover and credential stuffing, spam, malware, DDoS, and circumventing access controls on non-public data — anything behind a login, a paywall or an authorisation check. Collecting publicly accessible data is a legitimate use and the policy says so; you remain responsible for making sure your specific crawl is lawful, both where you are and where the site is.

Use case · Web scraping

Proxies for web scraping and data teams.

Self-serve: no sales call, no company-only onboarding, no plan that voids unused traffic at the end of the month. Buy gigabytes once, point Scrapy, requests, Crawlee or Playwright at one proxy URL, and move a domain from the ISP pool to a residential one by changing a single token.

Get started Jump to the snippets

Pay per GB · No subscription · Prices in EUR, excl. VAT

five domains, three poolsillustration

http://USERNAME{_pool-arc | _pool-arcfast | _pool-arcultra}:PASSWORD@proxy.arcproxies.io:8000

One gateway, one set of credentials. Moving a domain up a rail is one token in its username.

Illustration — fictional domains. Plan names and list prices per GB are the live catalog, excl. VAT.

€0.75
per GB on ARC Basic, excl. VAT: −21%
at 1 TB (€0.59 / GB): 185+
countries, city targeting: Never
expires — no subscription

The problem

Why crawls get expensive, or come back short

None of these are exotic. They are why a crawl that ran fine for a week starts costing more, returning less, or both.

Residential rates for pages that don't need them

Most of a crawl is HTML from sites that serve any well-behaved client. Paying residential rates for those pages is usually the biggest line on the bill. The ISP pool fetches them at €0.75 / GB; move a domain up only when it asks for it.

One rate limit stalls the run

When every request leaves from the same few addresses, a 429 on one domain holds up everything waiting behind it. A rotating pool spreads a domain's requests over many exits, and a backoff keeps you from burning through them.

Page 2 doesn't match page 1

Paginated listings, search results and filters often keep a cursor on the server and expect one visitor. Rotate the IP between pages and the ordering shifts, items repeat or the cursor dies. A sticky handle holds one exit for the whole walk.

A browser fetches megabytes to read kilobytes

A rendered page pulls scripts, styles, images, fonts and third-party tags, and none of them are the data — but all of them go through the proxy. Fetch the HTML directly where you can, and block the heavy resource types where you can't.

Retries that never back off

A 429 retried immediately usually earns another 429, and every attempt costs bytes. Back off exponentially, honour Retry-After, and move to the next pool on the retry rather than on the first request.

Locked out of the provider you'd pick

Some providers now onboard verified companies only, have dropped pay-as-you-go, or void unused bandwidth at the end of a cycle. A freelancer or a three-person team should not need a call to buy 20 GB.

Setup

Choose the pool per domain

Start every domain on ARC Basic. Move a domain up one rung when it asks for it — not the whole crawl. The ladder is a dictionary in your middleware, not a plan you commit to.

Plan	Role	Session	Targeting	Price
ARC Basic Rotating ISP	Default for every domain A rotating ISP pool at our lowest price per GB. High-volume HTML from lightly protected sites lives here, and that is most of a crawl.	Rotating	Country	€0.75 / GB €0.59 at 1 TB (−21%)
ARC Fast Rotating residential	When a domain gets rate-limited Residential exits. Move a domain here once its 429s outrun your backoff on ARC Basic. Give every paginated walk its own sticky handle.	Rotating, or sticky per listing	Country, optionally city	€2.00 / GB €1.12 at 1 TB (−44%)
ARC Ultra Premium residential	When success rate beats price Premium residential, for the few domains where a failed page costs you more than the price difference. It is the priciest of the three rotating pools, so route only those domains through it.	Sticky per listing	Country + city	€3.50 / GB €2.17 at 1 TB (−38%)

ARC Basic — default for every domain1 line

http://USERNAME_pool-arc_country-de:PASSWORD@proxy.arcproxies.io:8000

ARC Fast — when a domain gets rate-limited1 line

http://USERNAME_pool-arcfast_country-de_session-k3f9a1x2_sesstime-10:PASSWORD@proxy.arcproxies.io:8000

ARC Ultra — when success rate beats price1 line

http://USERNAME_pool-arcultra_country-de_city-berlin_session-k3f9a1x2_sesstime-10:PASSWORD@proxy.arcproxies.io:8000

Prices are list prices in EUR, excl. VAT. Volume discounts apply automatically at checkout — full ladder on the pricing page.

Sessions

Rotate for sweeps. Stick for anything with a cursor.

Both come from the same credentials. The difference is whether the username carries a handle — and the handle is yours to choose.

Rotating — no handle: Each new connection can get a different exit IP. Sitemap walks, detail pages by ID, anything where one request stands alone.
Sticky per listing: Add _session- with a fresh handle when a walk starts and pass it to every page of that walk. Set _sesstime- (minutes) longer than the walk.
Sticky per worker: One handle for a worker's whole run, when a short sequence of pages has to look like one visitor from start to finish.

Two workers, 12 requests each

worker-1

k3f9a1x2p4d8w1seh6c0n3ybAAAABBBBCCCC

worker-2

r2v7k5qam9t1f4zuDDDDDDEEEEEE

A new handle for every listing. All pages of one listing leave from the same exit; the next listing gets a new handle and a new IP. Right for pagination, sorted results and filters — anything with a cursor.

Letters stand for distinct exit IPs. Holding is best effort: an exit can go offline before its minutes are up, so a walk should tolerate a changed IP.

worker-1, first listing · ARC Fast1 line

http://USERNAME_pool-arcfast_country-de_session-k3f9a1x2_sesstime-10:PASSWORD@proxy.arcproxies.io:8000

Bandwidth

Billed per gigabyte, so fetch fewer of them.

Four habits move the bill more than the choice of pool does. The table puts illustrative numbers on them.

Fetch the HTML, don't render it. A browser is for pages that build their content in JavaScript; for everything else an HTTP client gets the same document for a fraction of the bytes.
When you do render, block images, fonts and media. It is one route handler — the Playwright and Puppeteer tabs below have it.
Keep compression on. Scrapy, requests, httpx and Go ask for it by default; curl needs --compressed. Stripping Accept-Encoding multiplies the bytes for nothing.
Retry with backoff, honour Retry-After, and climb a pool on the retry rather than on the first request. Every attempt is bytes you pay for.

How the page is fetched	Per page	Per 1,000 pages	ARC Basic	ARC Fast
Headless browser, nothing blocked Scripts, styles, images, fonts, media and third-party tags — every byte of it through the proxy.	2.5 MB	2.5 GB	€1.88	€5.00
Headless browser, images, fonts and media blocked One route handler. The page still renders; the bytes that carry no data are never requested.	800 KB	0.80 GB	€0.60	€1.60
HTML over plain HTTP, uncompressed requests, httpx, Scrapy, Go — the document only, with compression turned off.	300 KB	0.30 GB	€0.22	€0.60
HTML over plain HTTP, gzip or brotli The same document, compressed. Scrapy, requests, httpx, Go and Crawlee ask for it by default; curl needs `--compressed`.	60 KB	0.06 GB 42× smaller than a full render	€0.05	€0.12

Headless browser, nothing blocked
Scripts, styles, images, fonts, media and third-party tags — every byte of it through the proxy.
2.5 GB / 1,000 pages
Per page
2.5 MB per page
ARC Basic
€1.88
ARC Fast
€5.00
Headless browser, images, fonts and media blocked
One route handler. The page still renders; the bytes that carry no data are never requested.
0.80 GB / 1,000 pages
Per page
800 KB per page
ARC Basic
€0.60
ARC Fast
€1.60
HTML over plain HTTP, uncompressed
requests, httpx, Scrapy, Go — the document only, with compression turned off.
0.30 GB / 1,000 pages
Per page
300 KB per page
ARC Basic
€0.22
ARC Fast
€0.60
HTML over plain HTTP, gzip or brotli
The same document, compressed. Scrapy, requests, httpx, Go and Crawlee ask for it by default; curl needs --compressed.
0.06 GB / 1,000 pages
Per page
60 KB per page
ARC Basic
€0.05
ARC Fast
€0.12

Page weights are illustrative averages — measure yours per domain for a day. Euro figures are list prices per 1,000 pages, excl. VAT, before the volume ladder. Retries multiply every row: at a 10% retry rate, 1,000 pages is 1,100 fetches, and a retry sent without backoff usually fails again.

Code

One snippet per tool

Each one uses the tool's documented proxy option. Replace USERNAME and PASSWORD with the credentials from your dashboard; everything after USERNAME is targeting.

middlewares.py · spider · settings.py34 lines

# middlewares.py — a pool per domain, one rung up per retry, a handle per listingfrom urllib.parse import urlsplitLADDER = [                                # lowest price per GB first    "USERNAME_pool-arc_country-de",       # ARC Basic  €0.75 / GB    "USERNAME_pool-arcfast_country-de",   # ARC Fast   €2.00 / GB    "USERNAME_pool-arcultra_country-de",  # ARC Ultra  €3.50 / GB]START_RUNG = {"listings.example": 1}  # domains that skip ARC Basicclass ArcPoolMiddleware:    def process_request(self, request, spider):        host = urlsplit(request.url).hostname        rung = START_RUNG.get(host, 0) + request.meta.get("retry_times", 0)        user = LADDER[min(rung, len(LADDER) - 1)]        if handle := request.meta.get("arc_session"):   # same handle, same exit            user += f"_session-{handle}_sesstime-10"        request.meta["proxy"] = f"http://{user}:PASSWORD@proxy.arcproxies.io:8000"# spider — start each listing with meta={"arc_session": secrets.token_hex(4)},# then hand the same handle to every page of itdef parse_listing(self, response):    yield from self.parse_items(response)    if next_page := response.css("a[rel=next]::attr(href)").get():        yield response.follow(next_page, self.parse_listing,                              meta={"arc_session": response.meta["arc_session"]})# settings.pyDOWNLOADER_MIDDLEWARES = {"myproject.middlewares.ArcPoolMiddleware": 740}RETRY_TIMES = 2              # at most two climbs; 429 is already a retry codeAUTOTHROTTLE_ENABLED = True  # error responses can only lengthen the delay

The middleware runs before HttpProxyMiddleware (750), which replaces the Proxy-Authorization header whenever meta.proxy changes — so a retry can climb a rung cleanly. RetryMiddleware counts retry_times for you and already retries 429. AutoThrottle paces each domain by its latency, and error responses can only lengthen that delay, never shorten it.

Before you scale

Set the pool per domain in one place — a middleware or a proxy configuration — so moving a domain is a one-line change.
One handle per paginated walk; reuse it to return to the same exit.
Measure GB per 1,000 pages per domain for a day before you size a purchase.
Generate endpoints over the API once per run, not once per request. The limit is per key.
Keep the password and the API key in environment variables, not in the repo.

The whole API in one file: api.arcproxy.cloud/api/v1/llms.txt, generated from the live catalog.

Second provider

Run it next to what you already have.

Nothing here asks you to migrate. There is no commitment and no term, and the gigabytes you don't use this month are still there next month.

Failover

Keep your current provider on the first rung and put ARC on the next. Scrapy's retry_times climbs the ladder request by request; Crawlee's tieredProxyUrls does it per hostname and keeps probing the cheaper tier, so the failover only spends bandwidth while it is carrying traffic.

Split by domain

The domains that fetch fine from an ISP exit go to ARC Basic at €0.75 / GB and stop paying residential rates. The rest stay where they are. One dictionary entry per domain, one deploy, nothing migrated.

Canary

Send a fixed share of one domain through ARC for a week and compare success rate and GB per 1,000 pages with what you have. ARC Starter is 3 GB — 1 GB of each rotating pool — for €4.90, once per account.

Same line format as your list

The API returns lines as http://user:pass@host:port, host:port:user:pass or user:pass:host:port, over HTTP or SOCKS5 — whichever your rotation list already reads. Append them; nothing else changes.

Works with anything that takes a proxy URL

Scrapy
Python requests / httpx
Crawlee
Playwright
Puppeteer
Selenium
Node.js
Apify actors
Go net/http
curl

Illustration

What a million HTML pages a month costs

An illustration with round numbers — your pages, payloads and retry rate will differ. The arithmetic is the real price ladder.

Assumptions

Pages per month: 1,000,000
Average page, compressed HTML: 60 KB
Retries, with backoff: +10%
Bandwidth per month: ≈ 66 GB

On ARC Basic

€47.52per month, excl. VAT

66 GB bought in one purchase at €0.72 / GB (−4% volume rung, applied automatically).

The same 66 GB on ARC Fast: €105.60 at €1.60 / GB. Route only the jobs that need it through the pricier pool.

Compressed HTML over a plain HTTP client is the lightest thing you can fetch. Render the same pages in a browser with nothing blocked and, on the table's illustrative weights, the bandwidth is about 42 times as much — check that before you pick a tool, not after.

Bandwidth never expires, so a bigger single purchase is only ever cheaper per GB. See the ladder.

FAQ

Scraper questions

Something not covered? Write to contact@arcproxies.io.

Get started

Add a second provider in one deploy.

Create an account, and code WELCOME15 takes 15% off your first bandwidth purchase, up to 25 GB. No subscription, no sales call, and the gigabytes never expire.

Create your account Questions first? contact@arcproxies.io

AI training data

Collect public web text, images and product data at corpus scale from residential and ISP IPs, paid per gigabyte with no monthly commitment.

Alternative data

Turn public listings, pricing and hiring signals into a time series your models can trade on — collected on schedule, from the market you are modelling.

Real estate & jobs

Listings and postings are ranked by where the visitor is. Collect them city by city so your index reflects the market, not your server's location.

Price monitoring

See the price a shopper in Berlin, Lyon or Chicago sees. ISP bandwidth at our lowest price per GB for HTML checks, residential where a marketplace is strict.