arrow_back Back to Overview

Enterprise-Grade Security

Your business data is sensitive. We treat it with the respect it deserves.

lock

Encryption at Rest

All sensitive message content, including subjects and bodies, is encrypted using AES-256-GCM before it ever touches our database. Only our application, running within your secure environment, holds the keys to decrypt it.

shield_person

PII Scrubbing

We utilize a PII-scrubbed search index. This allows you to perform lightning-fast keyword searches across your inbox without exposing unencrypted Personally Identifiable Information (PII) to the search engine.

isolate

Tenant Isolation

Identity lookups are mailbox-isolated. Using a salted email_hash system, we ensure that contact data cannot be correlated across different mailboxes, maintaining strict privacy for your customers.

history_edu

OAuth 2.0 Native

Connect your Gmail or Outlook accounts using native OAuth 2.0 flows. We never see or store your primary account passwords—only the specific, revocable access tokens required to process your enquiries.

smart_toy

AI, On Your Terms

Reading and classifying your enquiries is done with our own proprietary logic, on our own infrastructure — that's true for every account, by default. Your message content is never sent to OpenAI, Anthropic, or any other language model provider to power the core product.

Want AI to help you write faster? Turn on AI reply drafting for a mailbox, any time. Once it's on, a “redraft with AI” button appears on every draft — one click sends that draft plus the visible email thread with that one contact to our AI provider (Anthropic) to return a rewritten version. Personal names, email addresses, phone numbers and postcodes are replaced with placeholders before that request leaves our servers, and restored locally in the draft that comes back. It's never automatic — a person always reviews and approves before anything sends — and no other mailbox, contact, or account data is ever included. The choice is entirely yours: leave the setting off, and nothing is ever sent.

Turn it on, and we also build a reply library for the mailbox — a distilled summary of how the business tends to answer common questions, refreshed about monthly from a sample of your own past replies. The same redaction applies: personal names, email addresses, phone numbers, postcodes and links are stripped before anything leaves our servers. The library is stored on our infrastructure, fully editable by the account owner, and used only as background reference for the redraft button — never sent to a customer as-is. Turn the setting off, and this stops too.

One more place a model is used, separate from your mailbox: before a bulk campaign goes out, you can run a compliance check. It reads only the campaign's own wording and the rough audience size, never anyone's emails or details, and flags anything that looks like it could breach marketing rules. It's guidance, not legal advice.

verified Technical Integrity

Our architecture follows the principle of Least Privilege. Background workers and API integrations are strictly scoped to the specific mailboxes and data points they need to function.

Shared mailboxes are restricted by default. A colleague you add to a mailbox sees only its enquiries and the conversations with the people who sent them, never your other mail, newsletters or invoices, unless you choose to give that person everything. The limit is applied on our server on every page and every action, and anything it does not explicitly allow is refused, so a missing check fails closed rather than open.

Every SQL query involving sensitive data utilizes prepared statements and deterministic hashing to prevent injection attacks and horizontal privilege escalation.