​
Two independent limits

Every request passes through two independent checks before it reaches a node:
  1. Rate limit — requests per second, enforced via a token bucket
  2. Bandwidth quota — bytes per calendar month, enforced against your tier’s cap
Either one can reject the request. The error codes are distinct (429 vs 402) so you can tell which.

​
Rate limit (requests per second)

TierRate limitBurst cap
Free3 rps30 tokens
Developer10 rps100 tokens
Team25 rps250 tokens
Scale50 rps500 tokens
Algorithm: token bucket. The bucket refills at your tier’s rate (e.g. 10 tokens/sec on Developer). Every CONNECT consumes 1 token. A full bucket lets you burst at 10× your sustained rate for ~10 seconds — useful for spiky workloads — then throttles back to steady-state. When the bucket is empty, the gateway returns HTTP 429 Too Many Requests. Back off exponentially and retry.

​
Bandwidth quota

TierIncludedOverage
Free500 MB / month— (gateway 429s instead of billing)
Developer5 GB / month$5 / GB
Team25 GB / month$4 / GB
Scale100 GB / month$3 / GB
Quota resets on the 1st of each calendar month (UTC). Exceeding returns HTTP 402 Payment Required.
Overage billing is opt-in per key. By default a key stops serving traffic when it hits its tier’s cap — no surprise charges. Enable overage from the /developer dashboard per-key.

​
What counts as “bandwidth”

Both upstream and downstream bytes through the tunnel. So a request that pulls 10 KB of HTML and POSTs 1 KB of form data counts as ~11 KB of quota consumption. The gateway is the canonical byte-counting authority (not the node). This matters: if a node provider self-reported their bandwidth usage, there’d be a financial incentive to over-report. The gateway sees every byte that flows through, so billing comes from there.

​
Monitoring your usage

Three surfaces:
  1. Dashboard. /developer shows current-period bandwidth used vs. tier cap, per key. Updated every 60 seconds.
  2. GET /api/v1/proxy/usage. Aggregate across all your keys. Good for programmatic check-ins.
  3. Alert email. We’ll email you at 80% and 95% of tier cap. Set preferred notification thresholds in account settings.

​
Scaling beyond Scale tier

The Scale tier (100 GB / 50 rps / 299)shouldcover>95299) should cover >95% of typical scraping workloads. If you're regularly hitting overage, email `sales@atlasvpn.live` — Enterprise custom contracts typically land between 500-2000/month depending on volume + country mix + dedicated node rental.