Graphical UI

Terminal Agent ships with a desktop Graphical UI (agent-gui) in addition to the agent CLI. It is a single, always-ready window that you summon with a global shortcut, type a prompt into, and dismiss again, without leaving whatever you were doing. It talks to the exact same providers, models, tools, and session logs as the CLI, so anything you configure for agent also applies here.

The window has two working modes that mirror the CLI commands: Ask for quick answers and Task for supervised, tool-using workflows. A left sidebar switches between those modes plus the History, Routine, Task Memory, and Settings screens.

Screens and features

For detailed information about each screen and feature, see:

Screen / feature	Description
Ask Mode	Quick questions with a streamed, rendered markdown answer
Task Mode	Supervised agentic workflow with live tool output
History	Browse recent ask, chat, and task runs
Routines	Manage scheduled, unattended agent runs
Task Memory	Browse and search what past tasks learned
Settings	Switch providers and models without editing a config file
Voice Input	Dictate a prompt instead of typing
Launching & Shortcuts	Install the global shortcut and use the keyboard controls

Appearance

The window uses a dark, terminal-native theme with monospace type and green accents. The lower-left shows the current working directory, and an animated mascot panel signals when the agent is thinking or busy.