Models and routers

With a Cloud account you can use every model in the model catalogue without signing up with each provider. Model requests are paid from your AI credits. If you already pay a provider, a router can use your own key instead.

Use Cloud models in Desktop and the CLI

1. Sign in: run haarness login, or choose Sign in to Haarness in Desktop.
2. In the model picker, choose Haarness Cloud and pick a model.

That is all. You do not need provider keys or an API key on your computer. Requests are paid from your plan's monthly AI credits. If you turn on Overage in Billing, purchased credits take over when those run out.

Call models from your own code

The router speaks the OpenAI API, so any OpenAI-compatible client works.

1. Open API keys and choose Create API key.
2. Pick the Models only preset. If you want, set a Monthly spending cap (credits) or choose Only selected models under Model access.
3. Copy the key; it is shown only once. Store it in your secret manager as HAARNESS_API_KEY.

import os
from openai import OpenAI

client = OpenAI(base_url="https://router.haarness.ai/v1", api_key=os.environ["HAARNESS_API_KEY"])
reply = client.chat.completions.create(
    model="openai/gpt-6-luna",
    messages=[{"role": "user", "content": "Say hello"}],
)
print(reply.choices[0].message.content)

Model names look like provider/model. The model catalogue lists every model with its price.

Routers: choose who pays for models

A router decides which models your account can use and how they are paid for. Every account has a Default router that pays with Cloud credits. Create more in Routers → Create router:

With Your provider keys, choose Provider keys on the router, pick the provider and choose Save key. Only providers with a key are available, and the router never falls back to Cloud credits. Provider keys are encrypted and never shown again. A router with your own keys still needs an active plan or trial on the account.

Limit available models, under Configure, restricts a router to the models you list, which is useful for a team or an app that should only use approved models.

Use a specific router

• In your code: each router has its own endpoint on the Routers page. Use it as the base URL instead of https://router.haarness.ai/v1. The standard endpoint always uses your default router.
• In Desktop: open the account menu and pick the Model router. It appears when your account has more than one router.
• In the CLI: list your routers with haarness account --routers, then choose one with haarness account --router ROUTER_ID.

Keep spending under control

• Per key: an API key can have a Monthly spending cap (credits), a Maximum credits per request, Requests per minute and Concurrent requests.
• For the whole account: turn Overage off in Billing. When your monthly AI credits run out, model requests stop instead of using purchased credits.

See What a model request costs for worked examples.

Run the router yourself

To run the model router on your own infrastructure with your own provider accounts, see Run the router yourself.