Security

Last updated July 15, 2026

Senttri connects to your email, so we treat security as a first-class product requirement, not an afterthought. Here's specifically how.

We never see your email password

For Gmail and Outlook/Microsoft 365, Senttri connects via OAuth. You authorize access directly through Google or Microsoft's own sign-in page — we receive a revocable access token, and your actual account password never passes through Senttri's servers at all.

Yahoo and iCloud don't support OAuth for third-party mail access, so those connections use an app-specific password — a separate credential you generate in your Yahoo/iCloud account settings, distinct from your real account password, which you can revoke independently at any time.

Encryption

  • All traffic to and from Senttri is encrypted in transit via TLS.
  • Our database and attachment storage are encrypted at rest.
  • OAuth tokens and IMAP app-passwords get an additional layer of application-level encryption (AES-256-GCM) before they're written to the database, using a key that's never stored alongside the data itself.

Access is scoped at the database, not just the app

Every table holding your mail, attachments, or team data enforces row-level security scoped to your organization's membership — meaning access control is a database-level guarantee, not something that can be accidentally bypassed by an application bug in a single API route.

What our AI sees — and doesn't

To categorize a message, we send its subject, a short preview snippet, and up to the first 500 characters of the body to our AI processing partner (Anthropic's Claude API) — never attachments. That data is not used to train models, by them or by us. If our AI processor is unavailable, categorization falls back to a local keyword classifier that never leaves our servers.

If you ask Senttri to draft a reply, more of the conversation is sent: up to the six most recent messages in that thread, each truncated to 1,200 characters. That only happens when you press "Draft reply" — drafting is never automatic. Attachments are still never sent. The AI that writes drafts has no ability to send, forward or delete anything; it returns text, and nothing leaves your inbox until you review it and confirm.

If you calibrate Senttri to write in your voice, we read a sample of your sent mail to measure style — typical greeting, sign-off, length and formality — and store only those descriptors. The sampled messages themselves are not copied or kept, and your voice profile is readable only by you, not by workspace admins.

Do you read my email?

No person at Senttri reads your email in the course of normal operation. Automated systems process it to provide the product (syncing, display, search, AI categorization); a human would only ever access specific message content in the narrow cases of a support request you've opened, or where required to investigate abuse or comply with the law.

Deleting your data

Disconnecting an email account deletes its messages, attachment files, and open-tracking data from our systems — not just the connection. See our Privacy Policy for full detail.

Compliance roadmap

Senttri is an early-stage product. We do not yet hold SOC 2 or CASA certification — both are on our roadmap as the company and customer base grow, and we'll update this page when either is achieved.

Report a vulnerability

If you believe you've found a security issue in Senttri, please email security@senttri.com with details — we take reports seriously and will respond directly.