Documentation
Security & limits
How the app protects you, and the speed bumps you might meet.
A few protections worth knowing about, in plain terms — plus the rate limits that occasionally show up as a "try again shortly".
How Bertram protects your account
- Your mailbox password never touches Bertram. Sign-in to Bertram is handled by a dedicated authentication provider, and mailbox access uses OAuth keys — held server-side, never displayed, with support for encryption at rest.
- Every request is authenticated, and your data is scoped to you — one user cannot read another's threads, drafts, or meetings.
- Email is rendered in a sandbox. Messages are sanitized before display and shown in an isolated frame, so a malicious email cannot run scripts or imitate a login form inside Bertram. Risky attachment types are forced to download rather than render.
- The AI treats email as data, not instructions. Message content is fenced as untrusted before it reaches the model, so a crafted email cannot steer your summaries or drafts.
Rate limits
To keep the service healthy for everyone, a few actions have speed limits. In normal use you will not notice them:
| Action | Limit |
|---|---|
| AI actions (summarize, draft, prioritize, chat…) | About 20 per minute |
| Autocomplete while typing | More generous — about 60 per minute |
| Sending email | About 10 per minute |
| AI actions per day | A daily allowance (500 by default), reset each day |
Hitting a limit is not an error in your account — the response tells you how long to wait, and the action works again after a short pause.