AttentionShield
· a calmer inbox
AttentionShield
· a calmer inbox
Checking your Gmail connection…
Nothing is processed while this is disconnected.
Automatically process new emails in your Gmail inbox
Included in your $9/month plan. We label mail Needs you / Can wait / Ads. We never hide mail until you say so.
Off by default. When a likely scam arrives, we send one short note to your own sign-in address: who it is from and why it looks wrong. Nothing else from the email. At most 3 a day.
These instructions are appended to the system prompt along with your other AI preferences.
Get a JSON file of everything we store about you: profile, preferences, email metadata and summaries, people, actions, feedback and billing. Never your Gmail tokens or message bodies.
Permanently remove your account and all associated data. This action cannot be undone. All your emails, summaries and preferences will be deleted immediately.
You never need these. The defaults work well; change them only if you want finer control.
Off by default: human-to-human email passes through untouched. Turn on to summarize everything.
Configure how the bot discovers new emails
How far back to search for unprocessed emails
Maximum number of emails to discover at once (1-100)
AttentionShield groups junk into packs (like “Sprouts ads”) and can hide them for you. You only approve — or let balanced mode learn after you clear a pack twice.
Never trashes mail. Default is label + skip inbox. Every rule can be undone in Shield.
Without this, AttentionShield still mutes senders and archives with labels using your existing Gmail access. Turning this on asks Google for mail settings permission so permanent filters can be created — the same power as Gmail Settings → Filters. You will re-connect Gmail once. We never trash mail by default.
Hiding a sender still lets their mail arrive. With this on, cleanup cards can also offer to use the sender's own one-click unsubscribe, so the volume drops where it's made. Still approve-first: you tap once to see who can be asked, and again to actually send it. It can't be taken back — we can't un-ask a sender to remove you, so there's no undo button and we won't pretend otherwise. Never offered for banks, receipts or security mail, and we never unsubscribe by sending email from your account.
When someone writes for the first time, AttentionShield asks you once whether to let them through. It only ever applies to mail that arrives after you turn this on — everyone already in your mailbox counts as known, so nobody you've been talking to for years gets flagged. Colleagues on your own domain and anyone you've replied to are never screened.
Hold unlocks after you've made a few decisions, so you can see what it does first. Anything you don't get to comes back to your inbox on its own — we release, we never quietly mute someone for you.
Privacy: rules use senders/domains from summaries, not full bodies. Audit of applied rules lives under Shield packs.
Who the shield leaves alone and who it hides. Changes apply immediately.
Mail passes through untouched — no summary.
Archived out of your inbox under the Muted label — never trashed.
Whole-domain mutes, usually installed by approved noise packs.
Only process emails longer than this many characters
Customize how AI generates email summaries
Every summary already includes action items, sentiment and a priority read — there's nothing to switch on.