Boring on purpose.

Access & authentication

• Sign-in is an email one-time code — short-lived, rate-limited, single-use. There are no passwords to phish or breach.
• AI connections use OAuth 2.1 with PKCE and dynamic client registration. Grants are per-user, revocable instantly at app.greentrees.ai/connect, and refresh-token reuse kills the whole grant on the assumption it leaked.
• Mailbox and calendar access uses Google’s and Microsoft’s own consent screens, with the narrowest scopes that do the job, revocable at the provider.

Isolation by construction

Every query runs scoped to your workspace at the database index level. There is no cross-tenant code path to forget to check — the shape of the query makes leakage a type error, not a code-review hope.

The AI is budgeted and receipted

• Per-connection hourly call and write ceilings starve a runaway loop or a prompt-injected write flood long before it can make a mess.
• Every write carries a receipt: what changed, who asked, over which channel, at what time — permanent, and undoable where undo is honest.
• Send-adjacent actions can never execute directly: they queue for a human. Approval links are signed, single-purpose, and expire in about three days; the AI is structurally unable to click them.

Data protection

• All traffic is TLS in transit; data is encrypted at rest by our infrastructure provider.
• Our infrastructure subprocessors (Convex, Vercel) maintain SOC 2 Type II programs; we build on their audited foundations.
• Access to production data is limited to the people who operate the service, for operation and support.
• Integration credentials are stored server-side, never exposed to clients, and are being moved to field-level encryption as an active hardening project.

Reply-stop & outbound safety

An inbound reply pauses a sequence in the same transaction that records it — there is no window where a prospect has replied and a follow-up still fires. Quiet hours and daily caps are enforced where sends execute, not in a UI that can be bypassed.

Reporting a vulnerability

If you find a security issue, email security@greentrees.ai. We read everything, we respond quickly, and we will not pursue good-faith research conducted without harming customer data or availability.