AI emails you don't have to edit

Marty VaughnAI Agents on YouTube

Just watched the local brain build? This is the kit from it. One prompt that reads your sent mail and builds a file on every person you deal with, plus the two skills that keep those files current and use them when you write.

Why AI emails usually need editing

It is not that the writing is bad. It is that the model has no idea who it is writing to. It does not know what you agreed with this person in March, what is still open, what they are like, or that the thing they are asking about is the thing you have been chasing them for. So it produces something polite and generic, and you rewrite it, which is most of the work you were trying to avoid.

The fix is not a better prompt. It is giving Claude somewhere to remember. This kit points it at a folder on your computer and has it build one file per person: who they are, what you have discussed, what was decided, what is still open. Next time you write to them it reads that file first.

You:  Help me write an email to Victor Lang asking if I can
      get status updates in a meeting tomorrow at 10:00 am.

Claude:
      Hi Victor,

      Are you free tomorrow at 10:00? I would like to get a
      status update on the thruster and the housing pricing.

      Thanks,
      Marty

      Heads up before you send: you already have an unsent
      draft to Victor proposing 9:00 tomorrow. I can delete
      it. Also, his file has an open question on whether
      Helix can hold the March 22 delivery date, and
      tomorrow is a good chance to get clarity on it.

What is in the kit

• The setup prompt. You paste it once. It asks how far back to scan, reads your sent mail, shows you everyone it found so you can cut people out of the list, and then builds the files.
• Two skills. One keeps the files current as you go, the other knows when to reach for them and what to pull out when you are writing. You install them once and they are always there.
• A setup guide. Four steps, written for somebody who has not done any of this before.

What you can do with it once it is built

• Reply to anything. "Reply to this." The draft comes back with the history already in it, and anything that did not belong in the email is shown to you underneath rather than stuffed into it.
• Ask about a person. "What is the latest with John?" It refreshes from your mail first, so you are never reading something stale.
• Keep your own notes in the loop. There is a folder for meeting notes, contracts, project dates, anything, and Claude reads whatever looks relevant when it drafts.

What you need

Claude on the desktop, on a paid plan, because the part that reads and writes files on your computer is Cowork and that is where it lives. Gmail connected to it, which takes a minute. Google Calendar is optional and only matters if you want Claude checking your availability when somebody asks to meet.

What it costs to run

Nothing beyond the Claude plan you are already paying for. There is no API key, no second account, and nothing to sign up for. The one-time build is the only expensive part in terms of usage, and it is expensive in minutes rather than dollars: ten to twenty, depending on how much mail it is reading. Scan thirty days the first time rather than ninety. It finishes far faster and it is enough to tell you whether you want this.