Connect a model
Tyrn doesn’t sell model access. You connect accounts and keys you already have, and your provider bills you. Connected models are available to the Tyrn router choice in the session form.
Three ways to connect
Open Settings and go to the INFERENCE group.
| Page | Use it for | Detail |
|---|---|---|
| Subscriptions | Sign in with an AI subscription, or add a coding plan key. | Use a subscription |
| Usage-based | API keys from providers that bill per token. | Use an API key |
| Local | Models on your machine or network, like Ollama or LM Studio. | Use a local model |
Native or Tyrn router
Each session chooses how its inference works, in the Inference field of the “New session” form:
- Native: “Native: uses
<Harness>’s own sign-in”. The harness talks to its own provider with its own account. You pick a Model and Effort in the form. - Tyrn router: the harness sends its requests through the providers you connected in Tyrn. You pick a model from a picker, and a Connect provider shortcut takes you to the settings if you have none yet.
Some harnesses can’t use the Tyrn router. The form tells you: “<Harness> does not support Tyrn router inference”, with the reason.
Choose a model per session
The model is part of each session, so you can run one task on one model and another task on a different one. Model ids appear exactly as the provider names them. In settings, a model is written "provider/model", or "provider:account/model" when you have several accounts for one provider.
Where credentials live
Provider keys and tokens are stored under ~/.tyrn/router/credentials/, in files readable only by your user account. Connector secrets and the Tyrn sign-in token are in the macOS Keychain.