Using the terminal UI¶
The terminal UI is a full-screen chat with your project: a transcript, tool rows, approvals, a composer and session history. It needs a terminal of at least 80 by 24 cells.
While tools run, the live indicator shows their activity. Successful tool receipts are hidden in native chat, including reopened sessions; the turn footer still counts them. Failures, unknown outcomes and approval requests stay visible. Plain text logs keep their existing tool receipts.
Start¶
ru chat in the current folder
ru agent chat --workspace C:\code\app
ru agent chat --preset plan read-only mode
ru agent run --workspace C:\code\app "explain the build" one prompt, then exit
ru with no arguments opens a chat in the folder you ran it from. ru agent run sends a single prompt
and streams the result; add --json for scripting. Plain text with no full-screen UI is available with
--classic, and Ru uses it automatically for redirected input or output, NO_COLOR, TERM=dumb or a
terminal smaller than 80 by 24.
Keys¶
| Key | Action |
|---|---|
Enter |
Send the prompt |
Ctrl+J, Shift+Enter, or \ then Enter |
New line in the prompt |
Ctrl+P |
Open the searchable command palette |
Esc, Ctrl+C |
Interrupt the running turn |
Ctrl+O |
Expand or collapse the model's reasoning block |
Ctrl+T |
Open or close the full Todos list |
Ctrl+Alt+Up / Ctrl+Alt+Down |
Select an earlier or later reasoning block |
| Drag across the transcript | Select visible text from one or more lines |
Ctrl+C with text selected |
Copy the highlighted transcript text |
Ctrl+Shift+C |
Copy highlighted transcript text, the selected reasoning block, or the latest assistant text |
PageUp / PageDown |
Scroll the transcript |
Ctrl+Home / Ctrl+End |
Jump to the oldest output, or back to following new output |
Up / Down |
Move through a multi-line draft, then through earlier prompts |
Ctrl+V |
Paste text, or attach a copied image |
While a turn is running, Enter queues your next prompt (up to eight). Queued prompts run in order.
Commands¶
Type / to see suggestions, or press Ctrl+P.
| Command | What it does |
|---|---|
/model [name] |
Switch or show the active model |
/provider [name] |
Choose a provider, then enter its API key |
/reasoning [effort] |
Show or set the active model's reasoning effort |
/advisor [off\|<model>\|<provider> <model>] |
Choose a model the agent may consult in this session (a new session has none) |
/computer [on\|off] |
Show or set computer use for new turns |
/status [overview\|context\|usage] |
Inspect status, context and token usage |
/preset [plan\|standard\|full_access\|rules] |
Choose permissions or manage remembered rules |
/new |
Start a new session |
/sessions [query\|id:<session-id>] |
Browse saved sessions and reopen one by id |
/subagents [run-id] |
See delegated tasks and read their results |
/tasks [run-id] / /tasks kill <execution-id> |
View and stop background terminal tasks |
/update [--yes] [--channel stable\|preview] |
Check for or install an update |
/help, /exit |
List every command, quit |
Custom commands you add appear in the same list; see Teach Ru your project.
Attachments¶
Attach images with --image PATH at startup (up to four), a dropped path or
Ctrl+V. Ru resizes and validates images before sending them; for a model that reads no images, an
image file is sent as a path instead. Any other file can be dropped too, whatever its type: text up to
8 KB is added to the prompt as a code block, and anything else (PDFs, Word or Excel files, archives,
large logs) is sent as a path the agent opens itself. A file from outside the workspace is first copied
into .ru\attachments, which git ignores. Backspace on an empty prompt removes the last attachment.
Attachments travel with the next prompt.
Sessions and paused runs¶
Each chat is a session. /sessions finds one and /sessions id:<session-id> reopens it, redrawing its
newest turns as they looked live. The native terminal shows recent saved messages first, then fills in activity and timing. A run that hits an error, for example a payment or quota problem,
ends with that error; send a new message to keep going once it is fixed. Typing Continue reopens a
run this session tracks that is still running or waiting on your approval or answer, and Escape
interrupts a running turn. A run that is still going is attached from another terminal with
ru agent attach <run-id> and abandoned with ru agent cancel <run-id>.
Deleting the conversation you have open returns to the empty landing instead of opening another saved conversation. Your next prompt starts a new session, so no empty conversation is created in between. While a turn is running, prompts are queued, or an approval/question is pending, Ru refuses a session or model change and says which of those is blocking it.
For a new prompt after idle shutdown, Ru starts or finds the Host and rechecks the selected provider and model before sending in the current session.
Todos¶
When Ru creates a checklist, the terminal shows completed, in-progress and pending items on the
right. A narrow terminal or a finished checklist shows a compact Todos 3/3 button. Click the
heading or button, or press Ctrl+T, to open the full list. Use the arrow keys or mouse wheel to
scroll, and Esc to close it. Reopening a session restores its latest recorded checklist.
More¶
The full command reference, limits and recovery rules are in the terminal UI reference.