Skip to main content
The Provider tab controls which AI model your assistant uses. A default model is preselected — fast, cost-effective, and included in every plan with no key or setup required.
Start with the default. Only switch if you have a specific reason to use a different model.

​
Choosing a Model

Model selection opens a full-screen picker, not a simple dropdown. It’s organised as a tab deck of reasoning categories — All, Frontier, Premium Reasoning, Economy, Specialized, Balanced Performance, Fast & Efficient — with a card for each model under the selected category. The picker currently offers three providers.
Only models from the default provider currently work when your assistant runs. The picker lets you select any model from any of the three providers, but selecting a model from either of the other two returns an error the first time the assistant is invoked. If you need to change models, stay within the default provider’s card set until this is resolved.
We deliberately don’t list model names here — the catalogue changes over time, and specific model IDs go stale within weeks. Pick from the categories in the picker itself; each card shows the model’s context window, output limit, and relative cost.

​
Bring Your Own Keys (BYOK)

BYOK lets you connect your own provider API key so your assistant calls the provider with your account instead of the shared default. It’s a flag on accounts with a Team plan (there is no Enterprise plan).
If your account doesn’t have BYOK enabled, the API-key section of the picker doesn’t appear at all — there’s no upsell shown in its place.
What BYOK actually does:
  • You get access to models and providers you couldn’t otherwise select
  • You never paste a raw key into this field — the control is a Select Key… dropdown over the API keys already stored in your account’s secrets, with an Add Secret option to add a new one
  • Only the secret’s name is stored on the assistant; the key itself is resolved at call time
BYOK calls are not free of QuivaWorks credits. A platform charge still applies on top of whatever your own provider bills you directly — BYOK removes the per-model markup, not the platform component.

​
Getting a Provider API Key

1

Create an account with the provider

Sign up on the provider’s own console. Most providers give new accounts free starting credit.
2

Set up billing

Add a payment method and set a monthly budget limit on the provider’s side before generating a key — usage on a BYOK key is billed by the provider directly, and QuivaWorks has no visibility into that spend.
3

Create and store the key

Generate a new key, copy it immediately (most providers won’t show it again), and store it in a password manager before adding it to QuivaWorks as a new secret.
Never share an API key or commit it to a code repository. If a key is exposed, revoke it on the provider’s console and issue a new one immediately.

​
Advanced Controls

Three additional controls exist but are hidden by default, behind two gates: Advanced Mode must be turned on globally (Settings → Advanced → Enable Advanced Mode), and this specific assistant’s Advanced Enabled toggle must also be on.
  • Temperature (0–2) — how deterministic vs. varied responses are
  • Max Turns (minimum 1) — how many back-and-forth messages are allowed in a single call
  • Max Output Tokens — capped by the selected model’s own output limit

​
Best Practices

The preselected model handles most work well and draws on your plan’s included credits. Only switch if you’ve hit a concrete limitation.
Selecting a model outside the default provider will fail at runtime today. If you’re testing a specific model from another provider, confirm it actually responds before building a flow around it.
Add provider keys through Add Secret in the picker rather than anywhere else. Use separate keys for development and production, and set spending limits on the provider’s own platform.

​
Next Steps

Context Settings

Configure memory, history limits, and output structure

Knowledge

Add documentation and reference sources

Tools & Connectors

Connect your systems via integrations

Best Practices

Optimise assistant performance and cost