SenseInbox 1.1 Beta · Gmail + Microsoft · reversible by design

Trust & safety

Powerful enough to help.
Restrained enough to trust.

SenseInbox is intentionally built around least privilege, protected-mail rules, explicit review, reversible provider actions and evidence-driven release gates.

No permanent mailbox deletion

The current SenseInbox provider surface does not implement a permanent-delete primitive. Cleanup uses reversible operations supported by Gmail and Microsoft providers.

Protected mail stays protected

Safety policy can retain high-value or low-confidence mail outside cleanup execution. Production pilot and certification gates explicitly test protected-mail invariants.

Review → preflight → execute → undo

Execution is separated from recommendation. Batches are preflighted against current provider state, recorded in an audit ledger and designed for recovery/undo.

OAuth and permissions

Mailbox connectivity uses provider OAuth. Organize/mutation authority is a separate capability from scanning and requires explicit provider permission.

Optional AI is not mailbox authority

External AI connectors are optional, can be disabled completely, use encrypted credentials when BYOK is configured, and receive privacy-reduced derived portfolio payloads. AI advice cannot execute mailbox mutations.

MCP is scoped and revocable

Remote MCP tokens are scoped, expiring and revocable. The current MCP surface contains read/planning/advisory tools only—no archive, trash, move, label, unsubscribe or automation mutation tool.

Operational evidence matters

Production rollout, provider parity, connector rollout, resilience and release certification are bound to exact source/environment/Worker provenance. Source tests do not stand in for live provider evidence.

Security questions or responsible disclosures: support@senseinbox.com