Every model,
one API key

Call frontier and open-weight models - text, image, video, speech, transcription, embedding, rerank and audio - through a single OpenAI-compatible endpoint. Swap providers by changing one string, keep your prompts, and let the gateway handle routing, automatic fallbacks, rate limits and per-team spend caps. Self-hosted and edge models join the same catalogue, so what runs in your VPC is called exactly like what runs in ours.