- Use a specific model that isn’t in the default catalog
- Run higher volumes without consuming workspace credits
- Take advantage of provider‑side pricing, rate limits, or compliance setup you already have
Supported providers
| Provider | What you get |
|---|---|
| OpenRouter | Access to OpenRouter’s full model catalog through one key |
| AI/ML API | Models served by AI/ML API |
| Fireworks AI | A curated set of Fireworks‑hosted open models, including Llama 3.1 405B, Llama 3.3 70B, Kimi 2.6, DeepSeek R1, DeepSeek V3, and Qwen3 235B |
Connect a provider
Paste your API key
Generate a key in the provider’s dashboard, paste it into the integration panel, and save. Builder verifies the key before storing it.
You can disconnect a provider at any time from the same Integrations screen. Disconnecting removes its models from the picker for everyone in the workspace.
How BYOK models behave
- Billing — Completions go directly to the provider on your key. They don’t deduct Native.Builder credits.
- Availability — If the key is invalid, revoked, or out of quota, BYOK models are hidden or fail at send time with the provider’s error.
- Switching models — You can switch between platform‑funded and BYOK models mid‑conversation; each message uses whichever model is currently selected.
- Plan gating — BYOK is available on every plan. The platform’s premium‑model gating only applies to platform‑funded models.
Next steps
Prompting best practices
Get better results from whichever model you pick.
Plans & credits
See which models are included on each plan.

