Quickstart

Mint a key and run a prompt on an Apple-attested Mac.

Mint a key and send a prompt. You get an answer from an Apple-attested Mac. If you already speak Chat Completions, that is a base URL and an API key.

Mifr will support many models. In this alpha the first ones are qwen3-4b and qwen3-8b. The examples on this page use qwen3-4b.

1. Get your API key

Sign in with Google at Settings and create an API key. The secret is shown once and starts with sk-mifr_.

Operator env keys (MIFR_GATEWAY_KEYS) are a different product; they are not minted here.

2. Make a request

Point the SDK you already use at the gateway. Origin:

https://mifr-gateway-production.up.railway.app

SDKs want that origin plus /v1.

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://mifr-gateway-production.up.railway.app/v1",
    api_key="YOUR_MIFR_API_KEY",
)

completion = client.chat.completions.create(
    model="qwen3-4b",
    messages=[{"role": "user", "content": "Reply with the single word pong."}],
)

print(completion.choices[0].message.content)
print(completion.usage)
print(getattr(completion, "mifr_receipt", None) or (completion.model_extra or {}).get("mifr_receipt"))

A 200 body is a chat.completion with usage and mifr_receipt. Verdict headers such as x-mifr-trust ride on the HTTP response; those headers are the gateway’s report, and mifr_receipt is the proof.

Streaming (stream=true) sends tokens as they arrive. The receipt and x_mifr_* verdicts land on the final chunk, not on the SSE HTTP head, because that head is flushed before a Mac is chosen. If the stream dies after the first token, there is no [DONE] and no receipt; start a new request instead of splicing onto those bytes.

A wrong key is HTTP 401:

{
  "error": {
    "message": "missing or invalid API key (Authorization: Bearer <key>)",
    "type": "invalid_request_error",
    "code": "invalid_api_key",
    "param": null,
    "mifr_action": "fix_key"
  }
}

mifr_action is the next step (fix_key, retry, wait, fix_request, pick_model). OpenAI SDKs ignore extra keys, so read status_code and the envelope.

Next steps

• Authentication covers how keys are minted, hashed, and rejected.
• Models covers aliases, the live list, and pinning a Mac.
• Chat completions is the live playground for POST /v1/chat/completions.