Skip to content

Add a Cloud Model Provider

The desktop ships a catalog of 34 providers, including Sakana AI (Fugu) as a first-class connectable provider. This guide covers connecting any of them in Settings > Models, whether by a single API key, keyless OAuth, or, for the cloud providers, a credential form.

The provider picker leads with real brand logos and a bring-your-own-endpoint section up front, and covers the full engine catalog of 104 providers. Flux Router is featured near the top of the providers screen. Every connected provider has a per-provider on/off toggle on its card in the grid: flip it off to stop the provider’s models from appearing in the model picker without deleting its saved credentials, then flip it back on later to re-enable it instantly.

Go to Settings > Models. The grid shows every provider in the catalog. Find the one you want and choose Manage.

xAI Grok supports a “Sign in with X” flow that connects without an API key. See Sign in with X (Grok) for the full walkthrough.

ChatGPT subscribers can connect their subscription without an API key using the “Sign in with ChatGPT” button on the ChatGPT card. This uses the same credential mechanism as the Codex CLI. See Sign in with ChatGPT for the full walkthrough.

Most providers connect with a single key:

  1. Paste the API key into the credential form.
  2. Save. The provider shows as connected.

The app recognizes many keys by their prefix and selects the right provider for you, so a pasted key often lands on the correct entry without you choosing it first. Recognition now covers GitHub Models tokens, OpenAI service-account and admin API keys, and the new Google AI Studio key format, for broader provider coverage.

AWS Bedrock, Google Vertex AI, and Azure OpenAI use a multi-field credential form rather than a single key, because they authenticate with cloud credentials. Open Manage for one of these and fill in the fields it asks for, such as region, project, and the credential it accepts.

After a provider connects, set a default model for it, or pick a model per conversation from the launch bar. You can also browse the provider’s available models and add a specific one.