The right model for each prompt.

We score your draft, optionally improve it, pick a model, stream the answer, then show follow-ups and an API snippet.