Drop-in compatibleFully compatible with the OpenAI SDK. Replace the base URL. Everything else stays the same.
Model routingOne API key, every model. Switch providers by changing a single parameter — no key rotation, no code changes.
Multi-channel failoverRequests are load-balanced across upstream channels. Failures trigger automatic retry on healthy backends.
Usage transparencyPer-request token counts, model attribution, and cost breakdowns. Every call is logged and auditable.