NOX API

OpenAI Compatible. Privately.

Already shipping on the OpenAI SDK? Change one base URL. Every request now routes through the encrypted Nox network — sealed in transit, computed in an enclave, never logged.

Read the docs

Zero-log network AES-256 in transit 40+ relay regions

request-runner

READY

POSTapi.nox.network/v1/chat/completions

1234567891011

{
  "model": "nox-llama-3.3-70b",
  "messages": [
    {
      "role": "user",
      "content": "Explain confidential inference in one sentence."
    }
  ],
  "stream": true,
  "temperature": 0.6
}

11 lines · application/json

Private pipeline

Encrypting

Sealing payload · AES-256-GCM

Routing

Private node · Frankfurt (fra-07)

Inference

Model responding

Response

Streaming tokens

Response

› Press Run to send an encrypted request.

DROP-IN

Three lines from public to private.

Keep your existing OpenAI client. Point base_url at Nox, drop in your key, and every call becomes confidential — no rewrite, no new SDK.

Same request & response shapes you already use
Streaming, tools, and JSON mode supported
Per-key scoping with zero retained prompts

from openai import OpenAI client = OpenAI(    base_url="https://api.nox.network/v1",    api_key="$NOX_API_KEY",) resp = client.chat.completions.create(    model="nox-llama-3.3-70b",    messages=[{"role": "user", "content": "Hello, Nox."}],) print(resp.choices[0].message.content)

ENDPOINTS

A familiar surface, privately served.

The routes you know from OpenAI — each one terminating inside the encrypted network instead of a vendor's data center.

POST/v1/chat/completionsStream chat & tool calls through the private network. live copy

POST/v1/images/generationsConfidential image synthesis — prompts never logged. copy

GET/v1/modelsList models available on the encrypted relay. copy

Ship your first private request.

Generate a key, swap the base URL, and you are running on a network that forgets every prompt the moment it answers.

Get an API key API reference