Skip to content
Kerry Morrison // BetterStory
← All Projects

Gmail CLI

A command-line tool for processing email with AI. Triage, summarize, and respond to email without leaving the terminal.

[View_Source]

The Problem

Email is a constant source of context switching. The typical workflow—open browser, scan inbox, decide what needs attention, respond, repeat—fragments focus and wastes time on low-value decisions.

Most "AI email" solutions add another layer on top of this: now you have an AI assistant and your email client and your todo list. More tools, more switching.

The Approach

Gmail CLI takes a different approach: bring email into the terminal where I already work, and use AI to handle the cognitive load of triage and response.

Key design principles:

  • Pull, don't push. No notifications. Check email deliberately.
  • AI-assisted triage. Automatically categorize messages by urgency and type.
  • Draft responses. Generate reply drafts that match my tone and context.
  • Keyboard-first. Everything accessible without touching the mouse.

How It Works

# Check what needs attention
gmail triage

# Get AI summary of a thread
gmail read <thread-id> --summarize

# Generate a response draft
gmail draft <thread-id> "decline politely, suggest next week"

# Send directly
gmail send <thread-id> --draft

The AI understands context from past conversations, calendar availability, and explicit preferences I've configured.

What I Learned

Building this taught me several things about AI integration:

  1. The interface matters as much as the AI. The intelligence is only useful if accessing it is frictionless.

  2. Personalization beats general capability. A model that knows my writing style and context produces better drafts than a more capable model without that context.

  3. Defaults are decisions. The triage categories, the response templates, the summary length—each choice shapes behavior.

Current Status

I use this daily. It's not polished enough for a public release, but the core functionality is solid and saves me roughly 30 minutes per day on email processing.

Future work:

  • Better calendar integration for scheduling responses
  • Learning from sent messages to improve draft quality
  • Potential web UI for mobile access