The AI chief of staff
for your inbox.

No. Triagd works on top of Gmail, Outlook, or any IMAP mailbox. You keep using the client you have — Triagd just labels, drafts, and files in the background.

Gmail (Google Workspace and personal), Microsoft Outlook (Microsoft 365 and outlook.com), and any IMAP/SMTP provider — Yahoo, iCloud, Fastmail, Proton Bridge, custom servers. All three providers run through a shared abstraction so the triage behavior is identical across them.

We run a continuous eval against a labeled fixture set and publish the live numbers in our changelog rather than baking a single marketing figure into this page. The metric we optimize for first is Tier-1 recall — how often we catch the truly critical email — because missing one is the most expensive error. Every override you make is also fed back as a personal rule that sharpens the model for you specifically.

Sender, subject, snippet, and first 1,000 characters of the body. Never attachments. Anthropic does not train on your data. Full email bodies are not permanently stored.

Only if you opt in, per scenario, with strict guardrails (never to new contacts, never on Tier 1, always a 60-second undo window). Default is: Triagd writes the draft, you press send.

Filters match on keywords. Triagd reads each email like a junior chief of staff — understanding intent, urgency, and context — and writes replies in your voice. SaneBox sorts. Triagd thinks.

Those replace your email client. Triagd runs alongside whatever client you already use, applies its work in your existing labels, and leaves you in control. Zero workflow change.

Yes — 7 days on every plan. We do collect a credit card up front (it reduces signup spam and means you don't get locked out of your settings at day 7), and you can cancel any time before the trial ends from Settings → Billing — you will not be charged.

One click in Settings → Danger zone. We purge all stored data — emails, classifications, briefings, rules, tokens — within 72 hours, and revoke OAuth access immediately.