All guides
Security
How ZYLX keeps your store safe
Encryption, read-only defaults, and human approval — explained simply.
3 min readEasy
Why this matters
You're trusting ZYLX with keys to your business. Here's exactly how they're protected, in plain language.
- 1
Credentials are encrypted and redacted
- Every API key you save is scrambled with AES-256-GCM encryption before it touches our database.
- After saving, the browser only ever sees the last 4 characters (like ••••1a2b) — the full key is never sent back.
- Application responses and support diagnostics pass through credential redaction. Safe support reports contain health, dates, counts, and typed errors—not bearer values or hashes.
- 2
Read-only until YOU say otherwise
- Every integration starts in read-only mode — ZYLX can look but not touch.
- We even recommend creating your Shopify app with only 'read' permissions at first.
- To apply any change, three separate switches must align: the server allows writes, your workspace setting allows writes, and you approved that specific change. Miss any one and nothing happens.
- 3
Actions follow capability and approval gates
- Read-only AI connections cannot make durable business changes.
- Action-capable tools still check the connection scope, current workspace role, provider permissions, autonomy policy, and required approval. Multi-business actions return an outcome per target instead of silently skipping denials.
- Where an action lifecycle is supported, suggestions, approvals, rejections, applications, and verification outcomes are recorded with timestamps.
- 4
Your data stays yours
- Your workspace data is isolated with database-level rules (Row Level Security) — other accounts cannot query it.
- Revoke an AI connection any time from Agent Servers. Removing one business from a connection takes effect immediately and refresh cannot restore it.