Isolation
- Every record is tenant-scoped; queries filter by tenant on every request.
- Inside a workspace, operational data is additionally store-scoped; cross-store ids return not-found.
- Store API keys are hashed (SHA-256) and bound to one store.
Credentials and transport
- Ad-platform tokens, courier credentials, and bot tokens are stored encrypted and never echoed back.
- Internal service traffic is HMAC-signed with timestamps and replay protection.
- Sensitive values are redacted from logs and event snapshots.
AI agent guardrails
- MCP access requires an active Diamond subscription and browser authorization through OAuth 2.1 with PKCE.
- The owner chooses one store or all tenant stores, plus read-only or operational access, on the consent screen.
- Write tools are allowlisted per scope; destructive tools require a fresh confirmation token bound to the exact action.
- All agent requests are logged with connection, tenant, store, tool, and outcome attribution.
Your data
Your store, product, order, and customer data belongs to you. We process it to run the service, and you can export or delete your account data — see the Privacy Policy for details.
Frequently asked questions
Can one of my stores read another store's data?
No. Operational data is store-scoped inside your workspace: cross-store ids return not-found, and a store API key is bound to exactly one store.
Who can see my courier and ad-platform credentials?
Nobody, including you, after entry: credentials are stored encrypted, shown only as configured/hint, redacted from logs, and never echoed back in full.
Is the AI agent a security risk?
The agent uses a revocable OAuth connection with explicit tenant, store, and permission scope. Write tools are allowlisted, destructive actions require confirmation, requests are rate-limited, and every call is audited.