Provider setup
Set up AiApiy reverse-proxy
One key for OpenAI / Anthropic / Gemini — stable from China.
What is AiApiy
AiApiy is a routing aggregator: it unifies a dozen upstream model APIs (OpenAI, Anthropic, Google Gemini, xAI, etc.) behind a single OpenAI-Responses-compatible endpoint. One key, one baseURL, and Kition can hit every major model.
Two main use cases: a stable low-latency entry point for regions where official APIs are flaky (China especially), and developers who want one invoice and one key. Kition has no commercial tie to AiApiy — it is just the OpenAI-compatible router we test against most.
Get a key and baseURL
- Log in to the AiApiy console → API Keys → New Key
- baseURL is typically
https://api.aiapiy.com/v1(region subdomains exist for .cn or eu) - Top up or attach Stripe — AiApiy is prepaid
- Check the Models page for available aliases (e.g.
gpt-5.1,claude-opus-4-7,gemini-2.5-pro)
Add it in Kition
Use the custom provider template. Pick wire responses — AiApiy exposes the OpenAI Responses shape even when forwarding to Anthropic upstream.
{
"name": "aiapiy",
"baseURL": "https://api.aiapiy.com/v1",
"apiKey": "ak-...",
"wire": "responses",
"defaultModel": "claude-sonnet-4-6"
}Auth header
AiApiy reuses OpenAI-style Authorization: Bearer <key> — no extra header. If your deployment uses a private baseURL, paste it verbatim — Kition performs no path rewriting.
curl https://api.aiapiy.com/v1/responses \
-H "Authorization: Bearer $AIAPIY_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"claude-sonnet-4-6","input":"ping"}'Model alias table
AiApiy uses aliases to flatten upstreams into the OpenAI shape. Below are common mappings — the live list lives in the console Models page.
gpt-5.1/gpt-5— forward to OpenAIclaude-opus-4-7/claude-sonnet-4-6— forward to Anthropicgemini-2.5-pro/gemini-2.5-flash— forward to Googlegrok-4— forward to xAI- Some upstreams expose a
-thinkingsuffix that enables reasoning — see the console
Common errors
401 Unauthorized— key typo / suspended / balance drained402 Payment Required— zero balance; top up404 model not found— alias typo or upstream removed429 too many requests— AiApiy or upstream rate limit; auto-retried502 Bad Gateway— transient upstream failure, usually back within 30s- Streaming stalls — check local proxy / DNS; AiApiy endpoint itself is SSE
When to use it
Good fit: based in mainland China, behind a corporate firewall blocking direct API access, or comparing many models on a single invoice.
Skip it if: you already have direct OpenAI or Anthropic accounts with stable connectivity — direct is always lower-latency and cheaper than going through any middle layer.