Skip to main content

​
Migrate from OpenAI SDK

Switching from OpenAI to Routor takes two minutes and three line changes. Your existing SDK, your existing patterns, your existing code. Nothing else moves.

​
Before and After

​
Node.js / TypeScript

Before: 
import OpenAI from "openai";

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

const response = await client.chat.completions.create({
  model:    "gpt-4o",
  messages: [{ role: "user", content: prompt }],
});
After: 
import OpenAI from "openai";

const client = new OpenAI({
  apiKey:  process.env.ROUTOR_API_KEY,    // change 1
  baseURL: "https://api.routor.ai/v1",   // change 2
});

const response = await client.chat.completions.create({
  model:    "auto",                        // change 3
  messages: [{ role: "user", content: prompt }],
});
Three changes. Everything else stays the same.

​
Python

Before: 
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}]
)
After: 
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["ROUTOR_API_KEY"],
    base_url="https://api.routor.ai/v1",
)

response = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": prompt}]
)

​
Environment Variable

Add your Routor key to .env: 
ROUTOR_API_KEY=sk-routor-YOUR_KEY_HERE
Keep your OPENAI_API_KEY if you use OpenAI directly anywhere else. They are completely separate.

​
What Changes in Practice

OpenAI directWith Routor
Model usedAlways GPT-5.5Right model for each prompt
CostFixed GPT-5.5 pricing25 to 70% lower on average
FailoverNoneAutomatic across 13 providers
model in response"gpt-4o"Actual model used
Response formatOpenAI formatIdentical OpenAI format

​
Gradual Rollout

You do not have to switch everything at once. Routor runs alongside direct OpenAI calls. They are independent. Keep critical features on a direct GPT-5.5 call while you test Routor on lower-stakes requests: 
const routorClient = new OpenAI({
  apiKey:  process.env.ROUTOR_API_KEY,
  baseURL: "https://api.routor.ai/v1",
});

// Regular requests go through Routor
const regularResponse = await routorClient.chat.completions.create({
  model:    "auto",
  messages: [...],
});

// High-stakes requests: force COMPLEX tier minimum
const criticalResponse = await routorClient.chat.completions.create({
  model:             "auto",
  routor_tier_floor: "COMPLEX",
  messages: [...],
});

​
Verifying the Switch Worked

Check the model field in the response body. It will show the actual model used instead of "auto": 
console.log(response.model);
// "kimi-k2.6" — the provider's bare model id

// Non-streaming responses also include a routor object with the full decision:
console.log((response as any).routor);
// { model: "moonshot/kimi-k2.6", tier: "LIGHT", profile: "auto", savingsPct: 83.2, ... }
Or check the Logs page in your dashboard to see routing decisions in real time.
TypeScript users: extra fields like routor_tier_floor are accepted by the API and passed through by the SDK at runtime, but the SDK’s types don’t know them — add // @ts-expect-error above the line or cast the params object.