What is Butter?
Butter is an OpenAI-compatible API proxy which forwards requests to inference providers and caches their responses. Upon repeat requests, Butter serves spoofed responses immediately and deterministically, bypassing wasteful and unreliable generations and guiding AI systems down proven paths. The cache is template-aware, meaning it can serve responses based on structural similarity rather than requiring exact matches.Warning: Butter is experimental and subject to breaking API changes (and
quirky template induction bugs). Please contact us
before considering production use.
Quickstart
Start here
Follow our short quickstart guide.