AI privacy, data retention, tenant isolation, compliance, and audit — everything you need to make a decision, in the open.
Comments and DMs you receive are sent to Anthropic’s Claude API to draft replies. We use Anthropic’s no-training (Enterprise) setting, so your data is never used to train Claude models.
You can opt out anytime: Settings → Notifications → “Don’t generate AI drafts.”
If your customer asks not to receive AI-assisted replies, we show a label saying so, per the EU AI Act.
Active accounts: we keep your interactions, drafts, and decisions.
After cancellation: your data is kept for 90 days so you can come back, then permanently deleted.
Deletion is automatic and runs daily.
Every data query is scoped to your account — no other merchant can ever see your data.
On Enterprise BYOK, your Anthropic key is stored and isolated per merchant.
Each merchant’s integrations use their own separate credentials — never shared across accounts.
Built for GDPR (EU), APPI (Japan), and CCPA (California).
Stripe webhooks are signature-verified with replay protection.
Telegram webhooks are verified with a constant-time secret check.
TLS 1.3, auto-provisioned by Cloudflare.
Every staff impersonation session auto-ends after 5 minutes, notifies you on Telegram instantly, and is logged.
Status page: /healthz.
Found a vulnerability? Email security@kizuki.smartrich.ai.
Source code: github.com/hanamizukikabukijou/kizuki (private; read access available per merchant).
Pro merchants and above can download all of their activity as JSON / CSV.
Contact trust@kizuki.smartrich.ai