From OpenAI

CleverRouter speaks the OpenAI wire format. For most apps, migration is two lines.

The minimal diff

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY!,
});

const res = await client.chat.completions.create({
  model: 'gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello.' }],
});

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.CLEVERROUTER_API_KEY!,
  baseURL: 'https://api.cleverouter.eu/v1',
});

const res = await client.chat.completions.create({
  model: 'mistral/mistral-small-3.2', // an EU-hosted equivalent
  messages: [{ role: 'user', content: 'Hello.' }],
});

Two lines: the baseURL, and the model id.

Optional: use the typed SDK

If you want autocomplete, typed errors, and direct helpers for embed / rerank / transcribe, add @cleverrouter/sdk:

bun add @cleverrouter/sdk@alpha

import { createCleverRouter } from '@cleverrouter/sdk';

const cr = createCleverRouter({
  apiKey: process.env.CLEVERROUTER_API_KEY!,
});

// Exactly the same OpenAI SDK, pre-configured against the gateway:
const res = await cr.openai().chat.completions.create({
  model: 'mistral/mistral-small-3.2',
  messages: [{ role: 'user', content: 'Hello.' }],
});

Model mapping cheat sheet

Browse the live catalog at cleverouter.eu/models.

What stays the same

• Streaming. SSE in OpenAI chunk format.
• Tool calls. OpenAI function-call shape.
• Vision. type: 'image_url' content parts.
• JSON mode. response_format: { type: 'json_object' }.
• Python SDK, LangChain, LlamaIndex. Anything that takes a base_url.

What changes

• API keys are CleverRouter-issued (cr_live_*), not OpenAI's.
• Model IDs follow <family>/<model> — no bare gpt-4o-mini.
• Zero data retention by default. If your existing audit logs rely on retained content from your AI provider, you'll need to do that logging yourself.
• EU-only. If a model isn't hosted in the EU, it's not in the catalog.

Related

• SDK overview
• Provider pinning
• Why EU-native