Instead of entering API keys separately for each project, you can configure model credentials once at the organization level and let projects inherit them. This keeps your secrets in one place and makes rotating keys much less painful.
Head to Organization Settings → Model Secrets, find the provider you want, and click the three-dot menu to enter your API key.
You can configure credentials for:
Model Providers — OpenAI, Anthropic, Gemini, X-AI, DeepSeek, Mistral, Perplexity
Cloud Providers — Amazon Bedrock, Vertex AI
LLM Gateways — Portkey, LiteLLM
Just because you’ve set up credentials here doesn’t mean projects will automatically use them. Each project needs to explicitly opt-in by toggling Inherit from Organization in their Evaluation Model settings.
This is intentional—it gives you flexibility. Some projects might need different models or credentials, while others can just inherit the org defaults.