Skip to main content
The AI agent can verify that a visitor controls an email address before using account-related tools. This is separate from operator login OTP and is stored as short-lived conversation metadata.

​
OTP flow

Current controls include a 10-minute expiry, a 60-second resend cooldown, a maximum of five failed attempts, hashed code storage, and removal of the stored hash after success.
Email control proves access to that mailbox at that moment. It does not prove legal identity, account ownership in another system, or authorization for irreversible actions.

​
Contact conflicts

AI updates can discover a different proposed value for an existing contact. Rather than silently overwriting trusted fields, the contact service creates conflict records. Agents can list pending conflicts and either accept or dismiss a proposed value.

​
Contact data operations

Agents can update contacts, add/edit/delete notes, manage tags individually or in bulk, delete selected contacts, and review conflicts. AI tools can seek or upsert contact context through secret-protected internal routes.

​
Privacy safeguards

  • Collect only fields enabled by widget policy.
  • Never log OTP codes or raw provider credentials.
  • Limit notes to support-relevant information.
  • Use verified backend fields, not model summaries, for sensitive actions.
  • Apply tenant scope to contact email/session uniqueness and every lookup.
Last modified on July 17, 2026