Hard numbers.
Not marketing fluff.

The proxy endpoints at api.claudeapi.cheap and claudeapi.cheap/api/proxy — every model route. Uptime is calculated monthly on a 30-day rolling window from public Instatus checks (1-minute intervals, multi-region). Scheduled maintenance is announced 24h in advance and excluded from the calculation; everything else counts.

If a calendar month's uptime drops below 99.5%, every Pro user with active balance during that month receives 1% of their net spend in that month as a balance credit, applied automatically by the 10th of the following month. No claim form, no paperwork — we run the report and credit accounts directly.

7 days no questions asked. Email support@claudeapi.cheap with your account email, we refund the remaining balance to the original payment method (crypto: we send back to the wallet you paid from). The $19 Pro plan upgrade is also refundable within the same 7 days if you haven't used Pro-tier rate limits.

24 hours on business days (Monday–Friday, UTC). Most replies inside 6 hours. Email + Telegram. A real human at the same email account handles every ticket — no AI auto-replies, no support tier shuffle. Pro users get priority queueing.

Some reseller competitors quietly route requests through quantized models (lower-precision weights that produce subtly worse responses) to cut their own costs. We do not. Every request hits the full-precision model — same numerical behavior as direct Anthropic / OpenAI / Google. We verify by running parity benchmarks weekly.

Nothing changes — balance never expires. We do not charge inactivity fees, deduct "maintenance," or sunset old accounts. Top up once, draw down on your own schedule.

We pre-fund upstream provider accounts in bulk before reselling — we never sell balance we don't already own. If our upstream provider were to disappear tomorrow, every dollar of customer balance is independently covered by funds we hold separately. This is explicit in our Terms of Service §11.