questions people ask
FAQ
Does my mail leave my machine?
Only as the Claude API calls your assistant sends itself, under your own API key, covered by Anthropic's API terms. Everything else runs locally on your own computer. No SaaS in the middle, no third party with mailbox access, no data stored on anyone's server for you.
Do I need to be a developer?
No. You need to be comfortable pasting a prompt into Claude Code and answering its questions. The master build prompt does a preflight check, builds everything, shows you dry-run results, and only schedules things after you approve them. The setup guide assumes zero prior knowledge.
What does it cost to run?
Roughly $2 to $10 a month in Claude API usage at a typical mail volume, on top of the Claude subscription you already have for Claude Code. Most calls hit the cheapest model; the expensive one only fires when a draft is actually needed. You can set a hard spend cap in the Anthropic console.
How long does setup take?
The Google Cloud part takes about 15 minutes with the guide. The first layer (triage + drafts) typically runs the same evening. The other four layers are add-ons you bolt on whenever you feel like it.
Does it work with Outlook? And Apple Mail?
Gmail and Microsoft 365 / Outlook.com are both supported; the setup guide has a walkthrough for each. Apple Mail is a mail client, not a provider: if your account underneath is Gmail or Microsoft 365, the drafts your assistant writes simply appear inside Apple Mail. Only pure iCloud mailboxes are out, there is no API worth building on. Honest note: the Gmail path is the battle-tested one; the Microsoft path is built by Claude Code from the same architecture.
Windows?
The assistant needs a machine that is regularly on, running macOS or Linux (launchd or cron). Windows users can run it under WSL, but the guide does not hold your hand there.
What if I get stuck?
The package includes a troubleshooting FAQ with the 14 issues people actually hit, including the obscure macOS one that silently kills scheduled jobs. There is no 1:1 support at this price, and the guide is written so you should not need it.
Isn't there a free open-source tool for this?
There is: Inbox Zero is a solid open-source app with similar features. Self-hosting it means running a Next.js app with Postgres and Redis (a developer weekend), and the easy route is their paid hosted plan, a subscription. This framework is a different job: you build your own assistant inside Claude Code, in your own account, in your own voice, extend it with any rule you want, and nobody can price-hike it later.
Couldn't I just prompt Claude Code to build this myself?
You could, and that is exactly why this is a framework and not an app. What you are buying is the map: the prompts that survived thousands of real mails, the label taxonomy that does not confuse the classifier, the signature gotcha, the scheduler trap that silently kills jobs on macOS. Reverse-engineering that takes three or four evenings. This is those evenings, already spent, for $49. If you would rather spend the evenings, the free rules above will get you started.
What about updates?
Free updates for the lifetime of the product line: every 1.x release is included. When the framework improves, you get an email and the new version appears behind your original Gumroad download link. No fixed schedule; updates ship when they are ready.
Refunds?
30 days, no questions asked, handled by Gumroad. If you read it and it is not for you, you get your money back.
