Taiwan-domiciled runtime AI firewall. Ingress 12-vector prompt-injection block; egress credentials/XSS/shell/JWT redaction; per-request D1 audit log (APAC, raw prompts never stored); KV-backed API keys.
Powered by the same engine you can audit your own prompts with for free:
npx prompt-defense-audit "your system prompt" · 0 deps · 5ms · MIT · source
Status: /health
Usage:
POST /v1/chat/completions
X-Quartz-API-Key: <your quartz key>
Authorization: Bearer <your OpenAI/Anthropic key>
Content-Type: application/json
{ ...standard OpenAI chat-completion body... }📊 Live dashboard: quartz.tw/usage (paste your key)
📥 Apply for closed beta: quartz.tw/apply
📦 Cloud source: github.com/ppcvote/quartz-cloud