---
title: One key for multiple plans
description: Configure a routing key in the dashboard, then select a plan with a header or request metadata.
---

A routing key lets your backend use one secret for several plans. Send `X-Zorveus-Key-Alias: pro` to use your Pro key's models and spending limit

For example, your Free key can allow $1/day and your Pro key $10/day. The routing key selects the right key for each customer

## Set up in the dashboard

1. Open **API keys** in your Business organization
2. Choose **Group existing keys** and select the app whose keys you want to group. To create the keys and routing key together, choose **Create keys together**
3. Assign each key a selector, such as `free` or `pro`. Configure each key's allowed models and spending limit in the dashboard
4. Choose a default key if you want requests without a selector to use it
5. Create the routing key and copy its `zrv_route_...` secret. Store it in your backend, as you would an inference key

## Use it for inference

Set the routing secret as your API key. Choose the selector from the customer's plan in your backend

```python
import os
from openai import OpenAI

client = OpenAI(
	api_key=os.environ["ZORVEUS_ROUTING_KEY"],
	base_url="https://api.zorveus.com/v1",
)

response = client.chat.completions.create(
	model="YOUR_ALLOWED_MODEL",
	messages=[{"role": "user", "content": "Hello"}],
	extra_headers={
		"X-Zorveus-Key-Alias": "pro",
		"X-Zorveus-External-User-ID": "customer_123",
	},
)
```

This request uses Pro's settings. Keep the same external user ID when the customer's plan changes

You can instead put the selector in the JSON body:

```json
{
	"model": "YOUR_ALLOWED_MODEL",
	"messages": [{"role": "user", "content": "Hello"}],
	"metadata": {"key_alias": "pro"}
}
```

With the OpenAI Python client, add the body selector through `extra_body={"metadata": {"key_alias": "pro"}}`

If you send both, `metadata.key_alias` takes precedence over `X-Zorveus-Key-Alias`. Model discovery and usage reads use the header

Omit both selectors to use the default. A blank, null, malformed, or unknown body selector fails without falling back to the header or default. An expired or revoked selected key also fails. Without a default, you must send a selector

Choose the plan in your backend. Do not let customers select a more generous plan themselves

## Change the customer's plan

Change the selector on their next request, for example from `free` to `pro`. To prorate an allowance increase, also send `X-Zorveus-Allowance-Change-Policy: prorate_increase`. See [prorated plan changes](/startups/prorated-plan-changes)

Manage the routes, default, spending limits, and key rotation in the dashboard

The OpenAI TypeScript client sends `metadata` directly in the request body. Its request options use `headers`, while the Python client uses `extra_headers`

## Use native SDK request options

The SDK does not create routing groups. Configure the group in the dashboard and set its secret as the inference API key. Your backend chooses the plan alias and customer ID

<CodeGroup titles={["TypeScript", "Python"]}>

```typescript
import { Zorveus } from "@zorveus/sdk";

const client = new Zorveus({ apiKey: process.env.ZORVEUS_ROUTING_KEY! });
const response = await client.chat.completions.create({
	model: selectedModelId,
	messages: [{ role: "user", content: "Hello" }],
	zorveusMetadata: { externalUserId: authenticatedUserId, keyAlias: "pro" },
});
```

```python
import os
from zorveus import Zorveus

client = Zorveus(api_key=os.environ["ZORVEUS_ROUTING_KEY"])
try:
	response = client.chat.completions.create(
		model=selected_model_id,
		messages=[{"role": "user", "content": "Hello"}],
		extra_headers={
			"X-Zorveus-External-User-ID": authenticated_user_id,
			"X-Zorveus-Key-Alias": "pro",
		},
	)
finally:
	client.close()
```

</CodeGroup>

For key-scoped discovery and usage, pass `headers: { "X-Zorveus-Key-Alias": "pro" }` in TypeScript request options, or `extra_headers={"X-Zorveus-Key-Alias": "pro"}` in Python. Do not change a shared client's customer identity between concurrent requests