UI Improvement: Better visual distinction between User Input and AI Output

Open 💬 6 comments Opened Jan 6, 2026 by ValadaresX

What feature would you like to see?

Current Behavior
The visual distinction between user input and AI output is too subtle.

  • User: Indicated by a small, dim grey > on the left.
  • AI: Indicated by a bullet point •.
  • Issue: Both use the same text color and font weight.

The Problem
When scrolling through the terminal history, it is difficult to quickly distinguish user prompts from AI responses. The current markers are low-contrast and hard to spot during a quick scan, requiring the user to focus intensely on the left margin to find context switches.

Visual Representation (Current State)
Everything looks uniform, like this:

> user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user
• ai response start, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response,...
> user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user, user
• ai response start, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response, ai response,...

(In the actual CLI, the icons > and • are very small and the text color is identical, making the separation invisible when scrolling)

Suggested Improvements
Please implement a better visual hierarchy. Possible solutions include:

  1. Color Differentiation: Use a distinct color for the User Input (or the AI Output).
  2. Separators: Add a horizontal line or divider between interactions.
  3. Backgrounds: Use a slight background tint for one of the actors.

Goal
To allow users to identify "who is speaking" instantly while scrolling, without relying on small pixel markers.

Additional information

_No response_

View original on GitHub ↗

6 Comments

VadimPushtaev · 7 months ago

The background is normally there, but it's missing for me in tmux e.g.

kornreaksmei-dev · 7 months ago

[[]()]()

MNeMoNiCuZ · 6 months ago

This should definitely be improved.

https://github.com/openai/codex/issues/10812

anand-shankar-1 · 3 months ago

This is a real problem - it's quite difficult to use compared to alternative agent harnesses. Please fix this 🙏

alexbi29 · 2 months ago

guys, please improve this: add an option to change background color on user inputs. Simple and easy.

aleksandar-radev · 4 days ago

+