Models & providers

Catalyst groups models by provider — Anthropic, OpenAI, Google, and so on. Each provider is a service, and a service offers one or more models. The AI Services screen (in the sidebar) is where you see everything available to you.

Built-in vs. your own

• Built-in models are provided by your workspace — the VooV Models service in the screenshot above. You can use them immediately, with no API key of your own. One is the default for new chats.
• Your own providers are services you add with your own API key — see Adding a provider. They sit alongside the built-in ones and work exactly the same way.

Either way, your API keys stay on the workspace and are never exposed in the browser — the model is called through Catalyst, not directly from your screen.

Switching models

The model in use is shown in the chat composer; open it to pick another. The same models are available to workflow LLM and agent nodes, so you can point different steps at different providers — even compare them in a single run.

Build the work, swap the model

Because nothing you build depends on a specific model, comparing providers costs you nothing but the run. Ask the same question of two models and keep the better answer.

Setting your default and pinning favorites

You can set a default model for new chats, and pin the handful you use most so they’re at the top of the picker. For built-in models you don’t own, your preferences (your default, your pins, your settings) are saved as your overrides — they don’t change the model for anyone else.

Next

• Adding a provider — bring your own OpenAI, Anthropic, Mistral, or one of ~20 supported providers, or a custom endpoint.
• Network devices — reach a model running on your own machine.