MorrowCache is a drop-in OpenAI-compatible proxy. Change baseURL—keep your model and API keys. Agents rephrase until done; each try is a full chat bill. Reuse cached completions when a same-intent judge matches (cloud Jev or local Kev/Laya)—not cosine alone. Demo: 2817ms → 423ms on a paraphrase HIT. Fail-open on judge errors. Freshness refuses stale "latest" hits. Streaming & tools bypass. npx @kushalicious/jevcache@latest start · morrowcache.vercel.app/agent-setup