Revolutionize your inbox.
Reclaim your time.

Any professional whose day lives in email. Founders, account managers, sales reps, customer success, operations leads, project managers, sourcing teams, real-estate brokers, lawyers, consultants — anyone who spends meaningful time reading and replying. The agent reads your inbox and adapts to how YOU work. Trading desks, agencies, B2B services, and ops-heavy teams have all been early adopters because email is their oxygen.

About 5 minutes to connect your mailbox (longer for CN providers — see the CN question below). Within 24 hours the deep-learning bootstrap reads your past correspondence and builds your voice profile + counterpart cards. After that the agent runs on a schedule (every 60/30/15 min depending on tier) — drafts sit in your Drafts folder by the time you open your laptop. The voice keeps sharpening from your edits, with a weekly refresh pass.

No. Hard rule — drafts only. Every email gets your eyes before it goes out. We request the read-and-draft permission from your mailbox provider but never the send permission. This is enforced at the API scope level, not just a setting.

No human at MerchGent AI ever opens your inbox. No customer success rep, no engineer, no operator. Your emails and drafted replies are processed entirely by machines, inside an automated pipeline. The only logs we keep are operational metadata (counts, lane breakdowns, run timings) — never email contents. If you cancel, we delete your stored mailbox tokens and voice profile within 7 days.

Mailbox credentials are stored AES-256-GCM encrypted at rest — the same authenticated-encryption standard banks use. Inbound emails are read in-flight via IMAP and never copied to our servers. The historical email corpus we read during the deep-learning bootstrap is processed in memory and discarded — only the derived profiles (counterpart cards, voice patterns, business summary) are persisted. Drafts live in YOUR own Drafts folder, on your mailbox provider's servers, not ours.

That's the whole point. MerchGent AI reads up to three months of your sent mail on day one and builds a voice profile from your actual writing. Drafts read like you, not like generic AI. And it gets sharper every week — every line you edit before sending teaches the agent.

Honest answer: drafts are usable from day one — they pick up your greeting style, sign-off, sentence length and vocabulary from the historical sample. But "sounds like the user wrote it" takes 2-4 weeks of normal use, because you'll edit each draft a little before sending and those edits teach the agent. Most customers report by week 4 they barely edit; by month 2 it's autopilot. High-volume senders converge faster (more signal); low-volume senders converge slower. The agent never stops learning — a weekly voice-refresh pass keeps it current with new patterns in your work.

No. The deep-learning bootstrap is hard-capped at ~1,500 of your most-recent sent messages plus a stratified sample of your top counterparts — high-volume senders don't blow the budget, and the sampling is diverse enough to capture your real voice. Ongoing usage is one classify-and-draft pass per inbound, also bounded. Our LLM provider is contracted not to train on the data, and the tier price is a flat monthly fee — no per-message surprises.

Three flat-rate tiers based on how often the agent runs against your inbox: Starter $29/mo (60-min ticks), Pro $59/mo (30-min ticks), Power $99/mo (15-min ticks + full-inbox sweeps). Every tier includes 1 mailbox; add extras at $19/mo each. Cancel anytime, no contracts. The 24-hour deep-learning bootstrap is included in your first month at no extra charge.

Your dashboard has an acceptance meter ("Your numbers") that counts how many MerchGent drafts you actually kept — i.e. sent, with or without edits. If that rate is below 50% across your second month of use, email support@merchgent.com from your account email and we'll make your third month free. No forms, no haggling — we read the same meter you do. And because the sent-matching is deliberately conservative (a draft only counts as kept when it can be confidently tied to a message you sent), the meter undercounts in your favor, never against you.

Microsoft 365 / Outlook (OAuth, one-click). Personal Outlook.com / Hotmail / Live (IMAP). Gmail and Google Workspace (IMAP via App Password — no OAuth required). All major Chinese enterprise mail: Tencent Exmail (腾讯企业邮箱), NetEase qiye (网易企业邮箱), Alibaba Cloud Mail, Sina. Plus iCloud, Yahoo, Titan, Zoho, Fastmail, and the universal "Any IMAP mailbox" option which covers Yandex, Hostinger, ProtonMail Bridge, self-hosted Dovecot, university mail, and anything else that speaks IMAP. If your provider doesn't appear, the generic IMAP path almost certainly works — just supply the IMAP host.

Honest answer: connecting Tencent Exmail (腾讯企业邮箱), NetEase qiye (网易企业邮箱) or Alibaba Cloud Mail is a two-step process that involves your IT admin. (1) Your domain administrator has to enable IMAP for your org in the provider's admin console — without this, the option is hidden in your personal settings. (2) You then generate a 'dedicated client password' or 'authorization code' (客户端专用密码 / 客户端授权码) in webmail — your regular login password will be rejected. We surface the exact step-by-step inside the connect flow, and we use the provider's overseas endpoints where they exist (e.g. Tencent's hwimap.exmail.qq.com) for faster connections from outside mainland China. None of the major CN providers offer OAuth for third-party mailbox access — so the auth-code dance is the path. Plan 5–10 minutes plus a quick word with your IT person.