Rethink the terminal: A portal to AI

Working with AI has fundamentally changed what I use a terminal for.

I now use it for almost all my work: coding, planning, review and drafting, from start to finish, and when revisiting it later. My focus is on understanding and steering AI’s work rather than running commands. That calls for fresh designs around this new kind of interaction, both inside the window and across the desktop.

So I have been building AgentTerm since January, and it has changed how I work (demo). These notes share the thinking behind its design.

Rethinking the terminal

Why I give each lead agent its own OS window and icon

Agents show up when needed.

Why I run AgentTerm from source

Works like an installed app. Personalize it without the friction.

Why documents, code and images open right in the agent’s terminal window

I can edit the rendered content to propose changes; my agent refines it and replies in place.

Building on the terminal’s text interface

AgentTerm understands agents; agents understand AgentTerm.

Building on the terminal

I have been adding capabilities on top of this new terminal.

Curated code review: explanations grounded in code

I removed a check so we focus on what needs my judgment.

A hub for agents across machines, with phone and voice access

AgentTerm annotates transcribed speech, and my agents correct garbled text using the context of our work.

---

AgentTerm has other features and design choices not covered above, such as commenting on agent output, copying it without reformatting, connecting to an IDE, running long jobs, coordinating a shared workspace, and setting up and invoking guides across AI CLIs using a namespace convention based on folder paths.

Use AgentTerm as is, or adapt it to your work.