Resell a Private Model

Use the Enterprise Customer Portal when your consultancy operates a model for multiple customers and wants customer-level authentication, limits, and usage without exposing the upstream provider credential. The request path is:

Customer application

The customer calls Planck’s OpenAI-compatible endpoint with its own sk-cp-... Planck key.

Planck gateway

Planck identifies the customer workspace, applies its limits and access policy, and records its usage separately.

Private upstream model

Planck routes the request through the consultancy-owned provider credential. The customer never receives that credential.

Fireworks private or fine-tuned models

Fireworks is supported as a normal BYOK provider. The model does not need to be in Planck’s public registry: an unknown model with an explicit provider route stays BYOK-only and is forwarded with the exact upstream model ID. In Enterprise → Customer Portal:

Create or open a customer.
Open Upstream model and connect a Fireworks API key owned by the consultancy.
Pin the exact model route. For example:

fireworks/accounts/your-account/models/customer-support-v2

Open Planck keys, create a customer key, and copy it once.
Give the Planck key and model route to the customer. Do not give them the Fireworks key.

The customer can then use the standard OpenAI client:

from openai import OpenAI

client = OpenAI(
    api_key="sk-cp-...",
    base_url="https://api.inquantum.ai/v1",
)

response = client.chat.completions.create(
    model="fireworks/accounts/your-account/models/customer-support-v2",
    messages=[{"role": "user", "content": "Hello"}],
)

Customer Portal keys do not fall back to Planck-through-billing. They use the upstream credential assigned by the consultancy, which keeps costs and provider ownership under the consultancy’s control.

What is isolated per customer

Planck API keys and revocation
Request and cost usage
Monthly request and cost limits
Members and workspace access
Model-access policies when configured

The upstream provider credential remains in the parent consultancy workspace and can be assigned without exposing its raw value.

The portal workflow covers supported providers such as Fireworks using their exact private model IDs. An arbitrary self-hosted base URL uses Planck’s separate custom-target allowlist and is not configured through this portal.

Getting Started

AI Gateway

Observability & Analytics

Prompt Management

Async Logging

Custom Logs

Resell a Private Model

Fireworks private or fine-tuned models

What is isolated per customer

​Fireworks private or fine-tuned models

​What is isolated per customer

Fireworks private or fine-tuned models

What is isolated per customer