Blog
Engineering notes from the people building ApiLink. Billing edge cases, integration guides, post-mortems — written so the next person doesn’t have to re-derive them.
Pay with crypto, get 5% more balance
Pay $10 in crypto, get $10.50 of credit. No coupon, no catch — just how it works and why crypto is now the cheaper way to top up.
424 models are now live on ApiLink
From a curated shortlist to 424 models across every major provider, synced automatically — same API, same key, way more choice.
How to choose an AI API gateway (a checklist, not a sales pitch)
Every gateway claims to be the obvious choice. Here's a checklist to evaluate any of them — including us — on your actual constraints instead of marketing copy.
The best OpenRouter alternative for teams in China
Same models as OpenRouter, synced automatically — plus RMB billing and real VAT invoices for China-based teams that OpenRouter structurally can't offer.
ApiLink vs OpenRouter vs ZenMux: an honest gateway comparison
Three AI gateways, side by side. Where each one wins, where each one loses, and the honest answer about using more than one.
Pointing OpenAI Codex CLI at a third-party gateway
Two environment variables and Codex talks to Claude, Gemini, or DeepSeek instead of GPT. Plus the same trick for Cursor, Aider, Cline.
Using Claude/GPT/Gemini from China: a compliance checklist
Payment, invoicing, forex, data residency — every wall Chinese teams hit a quarter into using OpenAI or Anthropic, with a concrete checklist.
Anthropic prompt cache: when the 25% write premium actually pays back
Anthropic's prompt cache is 90% cheaper — on reads. The catch is the 25% write premium. Here's the math on when that pays off.
Where streaming AI gateways quietly leak money on every disconnect
Most gateways refund users when streams drop mid-flight. A specific disconnect pattern turns that refund into free inference. Here is the attack and the fix.