Product
Bring your own model to Pilot
Pilot isn't tied to one model provider. A platform admin can point it at the SDK LLM Gateway, Kilo, or any OpenAI-compatible endpoint.
Model choice is an admin setting, not a hardcoded default
From the admin panel, a platform admin can add any OpenAI-compatible model gateway (the SDK LLM Gateway, Kilo, or a custom endpoint) with its own API key and an allow-list of models. Organizations then choose from whatever the admin has made available.
Why this matters
Pricing, latency, and model quality all shift over time. Pointing Pilot at a different gateway is a settings change, not a redeploy. Nothing about a gateway's identity is hardcoded into the product.