What you get: Free shared API keys to access models like Kimi, Qwen, and GLM via a unified endpoint.
Start here:
Watch out: Public keys deplete very quickly, causing immediate HTTP 401 unauthorized errors once the shared balance is exhausted.Four live models on one link, homework plus scripts handled today, key pasted below, fastest readers win 
I tested this key myself โ each call below answered live.
MY FREE-AI RECEIPT โ $500 LOADED, YOU PAY $0
ITEM 01 โ TALK HELP
Straight answers for questions, drafts, study notes.
ITEM 02 โ CODE HELP
Bug fixes, fresh scripts, full weekend bots.
Model line: kimi-k3 for talk, kimi-k2.7-code for builds.
ITEM 03 โ HEAVY HELP
Long reads plus hard sums stay on track.
Each model reads around 1M tokens per call.
Model line: qwen3.8-max for deep work, glm-5.3-flash for speed.
| Slip | Detail |
|---|---|
| Models on one key | 4, listed above |
| Reads per call | around 1M tokens |
| Formats live | OpenAI plus Anthropic |
| Price at checkout | $0 |
One link for all four, paste-ready:
api.iceyyy.dev/v1
MY KEY โ PASTE AS-IS
I keep this key public, spend away:
sk-krishn-2u2c4mkwndl5e5gl0y8g33hm320uzv8shqotqromv8lkbv0x
๐ ๏ธ Plug it in โ 60 seconds, curl trio
I point my old setup at my link.
I swap the model name, same key, same endpoint.
I swap MYKEY for my key above in each sample.
curl, OpenAI format, paste, run:
curl -H "Authorization: Bearer MYKEY" -H "Content-Type: application/json" -d '{"model":"kimi-k3","messages":[{"role":"user","content":"ship it"}],"stream":true}' https://api.iceyyy.dev/v1/chat/completions
curl, Anthropic format, same key:
curl -H "Authorization: Bearer MYKEY" -H "anthropic-version: 2023-06-01" -H "Content-Type: application/json" -d '{"model":"kimi-k3","max_tokens":256,"messages":[{"role":"user","content":"hello"}]}' https://api.iceyyy.dev/v1/messages
I send auth two ways, Bearer plus X-API-Key:
curl -H "X-API-Key: MYKEY" -H "Content-Type: application/json" -d '{"model":"kimi-k3","messages":[{"role":"user","content":"hello"}]}' https://api.iceyyy.dev/v1/chat/completions
๐ ๏ธ Plug it in โ SDK plus editor setups
I run OpenAI SDK: change base URL plus model name.
I run Anthropic SDK: change base URL plus model name.
I run LangChain, LlamaIndex, Vercel: change base URL.
I run Claude Code: set three env lines below.
Whatever I run, I point it here.
OpenAI Python setup:
from openai import OpenAI
client = OpenAI(api_key="MYKEY", base_url="https://api.iceyyy.dev/v1")
resp = client.chat.completions.create(model="kimi-k3", messages=[{"role": "user", "content": "explain monads, make it funny"}], stream=True)
for chunk in resp:
print(chunk.choices[0].delta.content or "", end="", flush=True)
OpenAI Node setup:
import OpenAI from "openai";
const client = new OpenAI({ apiKey: process.env.KIMI_KEY, baseURL: "https://api.iceyyy.dev/v1" });
const stream = await client.chat.completions.create({ model: "kimi-k3", messages: [{ role: "user", content: "write a haiku about race conditions" }], stream: true });
for await (const chunk of stream) {
process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}
Anthropic Python setup:
from anthropic import Anthropic
client = Anthropic(api_key="MYKEY", base_url="https://api.iceyyy.dev")
msg = client.messages.create(model="kimi-k3", max_tokens=1024, messages=[{"role": "user", "content": "refactor this like you mean it"}])
print(msg.content[0].text)
Claude Code setup, three lines:
export ANTHROPIC_BASE_URL="https://api.iceyyy.dev"
export ANTHROPIC_AUTH_TOKEN="MYKEY"
export ANTHROPIC_MODEL="kimi-k3"
I make these stick across sessions:
echo 'export ANTHROPIC_BASE_URL="https://api.iceyyy.dev"' >> ~/.zshrc
echo 'export ANTHROPIC_AUTH_TOKEN="MYKEY"' >> ~/.zshrc
echo 'export ANTHROPIC_MODEL="kimi-k3"' >> ~/.zshrc
source ~/.zshrc
Vercel AI SDK setup:
import { createOpenAI } from "@ai-sdk/openai";
import { streamText } from "ai";
const kimi = createOpenAI({ apiKey: process.env.KIMI_KEY, baseURL: "https://api.iceyyy.dev/v1" });
const { textStream } = await streamText({ model: kimi("kimi-k3"), prompt: "build me a landing page plus fresh style" });
Continue, Cline, Roo VS Code setup:
{
"models": [
{
"title": "Kimi K3",
"provider": "openai",
"model": "kimi-k3",
"apiBase": "https://api.iceyyy.dev/v1",
"apiKey": "MYKEY"
}
]
}
STREAMING, LIVE
I set stream true, words arrive one by one as made.
Example stream, live reply:
data: {"choices":[{"delta":{"content":"ship"}}]}
data: {"choices":[{"delta":{"content":" it"}}]}
data: [DONE]
PROOF I RAN IT
I sent hello through each format, each format replied hello back.
SMALL PRINT, HONESTI resend past messages to continue a talk.
I hit 429, I slow down, I wait, I retry.
This key stays public, shared pool, fastest readers spend first.
I use this key for test-drives, weekend builds.
Serious work belongs on your own key.
Errors arrive clean JSON, type plus message plus code inside.
GO BUILD, MY LIST
I ship agents.
I ship chatbots.
I ship a CLI roasting my commit messages.
I ship a Discord bot asking zero permission.
I stretch $500 further than you think.
TOTAL DUE: $0 โ GO BUILD
One link again before you scroll:
:
!