Skip to main content
The models command surfaces which LLMs the agent builder is allowed to select in this workspace. It’s the same filter the app applies: only providers with a key set (via nora providers) contribute models to the list. Set a provider key and the list grows. Nothing new — same catalog every workspace sees.

​
Commands

​
models list

Prints one row per available model — model name, provider, env-var name, and (if present) the custom API base URL. Add --json for the full response including per-model capability blobs:
Useful when piping into another script or an LLM.

​
Empty list

If the command prints “no models available”, you haven’t keyed any provider yet:
The BYOK filter is enforced server-side — same rule that hides unkeyed providers in the agent builder.

​
Custom OpenAI-compatible endpoints

When a provider is set with a custom api_base (self-hosted vLLM, LM Studio, an OpenAI-compatible proxy), the extra column shows the base URL so you can tell hosted from self-hosted at a glance.