Auto Router is delivered through the Kilo AI Gateway, a unified OpenAI-compatible endpoint that aggregates hundreds of models from providers such as Anthropic, OpenAI, Google, xAI, Mistral, and others behind a single API key. Rather than acting as a single trained model, it functions as a routing layer that selects an appropriate upstream model for each request, which fits the pattern of OpenRouter-style auto-routing keys exposed by aggregator gateways. The gateway is designed to drop in wherever an OpenAI client is already used, so developers can switch or auto-route models by changing a single model string in their existing code.
In practical terms, Auto Router inherits the gateway-level capabilities documented for Kilo's infrastructure, including Server-Sent Events streaming with time-to-first-token tracking, Bring-Your-Own-Key support with encrypted credential storage, per-request usage and cost tracking, organization-level model allow lists and spending limits, and robust function or tool calling with automatic repair for deduplication and orphan cleanup. The gateway also documents fill-in-the-middle code completions and compatibility with the Vercel AI SDK, the OpenAI SDK, and any other OpenAI-compatible client in any language, making Auto Router a flexible choice for applications that want to fan out to multiple upstream models without managing separate provider integrations.