Confidential AI API: Secure LLM Inference at Scale

Beta Router is a beta service that provides access to a curated model catalog through an OpenAI-compatible API.

Router currently exposes a broader chat-model catalog than the Confidential API catalog documented on this site. The tradeoff is explicit: Router requests are not confidential. Use the Confidential API for sensitive data and use Router when model breadth is the priority.

Router is not confidential. Do not send secrets, personal data, regulated data, or other sensitive data. Use Prem API for confidential processing.

Router uses the Chat Completions API. All models accept text. Selected models accept base64 images. The model catalog and API behavior can change during the beta.

Request an API key

Contact us to request a Router API key. Include Router access in your message. Store the key in an environment variable. Do not commit it to source control.

export PREM_ROUTER_API_KEY="your-api-key"

The Router key is separate from the Confidential API key. PREM_ROUTER_API_KEY and PREM_API_KEY are not interchangeable.

Make your first request

Python

Install the OpenAI Python SDK:

pip install openai

Create a Python file with this code:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["PREM_ROUTER_API_KEY"],
    base_url="https://router.prem.io/v1",
)

response = client.chat.completions.create(
    model="kimi-k3",
    messages=[
        {
            "role": "user",
            "content": "What can you help me build?",
        }
    ],
)

print(response.choices[0].message.content)

Run the file. The response appears in your terminal.

Node.js test

Use Node.js 22 or later. Install the OpenAI JavaScript SDK:

npm install openai

Create a file named router.test.mjs:

import assert from "node:assert/strict";
import test from "node:test";
import OpenAI from "openai";

const apiKey = process.env.PREM_ROUTER_API_KEY;
if (!apiKey) {
  throw new Error("Set PREM_ROUTER_API_KEY before you run this test.");
}

const client = new OpenAI({
  apiKey,
  baseURL: "https://router.prem.io/v1",
  maxRetries: 0,
  timeout: 120_000,
});

test("Router returns a chat completion", async () => {
  const response = await client.chat.completions.create({
    model: "kimi-k3",
    messages: [
      {
        role: "user",
        content: "Reply with exactly: Router is working",
      },
    ],
  });

  assert.equal(response.model, "kimi-k3");
  assert.equal(response.choices[0].message.role, "assistant");
  assert.match(response.choices[0].message.content ?? "", /\S/);
});

Run the live test:

node --test router.test.mjs

Models

Model availability, context windows, capabilities, and streaming requirements can differ by API key. Use authenticated GET /v1/models as the source of truth, then send an exact model ID from that response in your chat request. Router currently exposes a broader chat-model catalog than the Confidential API. This comparison is about catalog breadth, not confidentiality or model quality. A model listed for Router is not evidence that the same model runs in a confidential enclave. See Models for the base catalog, additional key-dependent models, modality support, and streaming requirements.

Models

Browse model IDs and check which models are available to your API key.

API capabilities

Use tool calls, structured output, reasoning, vision, log probabilities, and usage reporting.

Add a custom provider

Configure Router in OpenCode, Pi, and other OpenAI-compatible tools.

Get Started

Router

Core Concepts

Handling Errors

Guides

API Reference

Billing

Support

Router

Request an API key

Make your first request

Python

Node.js test

Models

Models

API capabilities

Add a custom provider

​Request an API key

​Make your first request

​Python

​Node.js test

​Models

Models

API capabilities

Add a custom provider

Request an API key

Make your first request

Python

Node.js test

Models