Enterprise AI teams commonly build LLM gateways to centralize routing, policy enforcement, and spend tracking. Snowflake Cortex Inference integrates into this pattern by exposing OpenAI-compatible and Anthropic-compatible REST APIs, allowing teams to add Cortex as a backend without changing application-side clients. LiteLLM now officially supports Snowflake Cortex as a provider, enabling PAT-based auth, streaming, tool calling, and embeddings. Key benefits include reduced credential sprawl via a single Programmatic Access Token, inference running within the Snowflake security perimeter with RBAC governance, and AI spend visibility through SQL in the same platform used for other Snowflake cost tracking. A practical getting-started guide covers generating a PAT, configuring the gateway base URL, and testing with a raw curl request.
