Zorveus LogoZorveusDOCS
Getting Started

Bring your own key (BYOK) quickstart

View Markdown

Connect an upstream provider credential and route eligible inference through it.

Bring Your Own Key (BYOK) allows your organization to store encrypted provider credentials in Zorveus. Requests matching your provider policy route directly through your upstream account.

Optional key for interactive API requests. Stored in memory only.


1. Add a provider credential

Add provider credentials under Provider Credentials in the dashboard. Credential management is not a supported service-key operation.

Navigate to Provider Credentials

In app.zorveus.com, go to Settings -> Provider Credentials.

Select Provider and Enter Key

Choose your provider (e.g. openai, anthropic, gemini, openrouter) and paste your raw API secret (e.g. sk-...).

Configure Model Policy

Set the model policy to determine which models resolve to this key:

  • *: Matches all models allowed for the credential
  • openai/*: Resolves all OpenAI models
  • openai/gpt-4.1-mini: Resolves only this specific model

2. Route a request

After saving the credential, create or update an app connection whose model policy permits the route. Credentials in auto_resolve mode can be selected automatically. Credentials in manual_grant_only mode also need an explicit grant to the connection.

Client SDK / Framework:
from openai import OpenAI

client = OpenAI(
    base_url="https://api.zorveus.com/v1",
    api_key="zrv_your_api_key",
)

# The route uses BYOK when an eligible OpenAI credential resolves.
response = client.chat.completions.create(
    model="openai/gpt-4.1-mini",
    messages=[
        {"role": "user", "content": "Explain the advantages of vector databases."}
    ],
)

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

3. Verify BYOK usage accounting

In your Zorveus Dashboard under Usage Analytics:

  • Billing mode: Tagged as byok.
  • Sell cost: The platform fee charged by Zorveus. $0.00 for the first $25,000/month in BYOK traffic. Volume above $25k incurs a configured platform routing fee.
  • Provider cost: The estimated upstream model cost (e.g. OpenAI or Anthropic rates) recorded for analytics.
  • Tokens: Token counts measured by the gateway or returned by upstream providers.

Next steps

Was this page helpful?
Edit this page on GitHub

On this page