v1.96.0 ·
The console shows everything and takes one key
Two days ago aidlc with no arguments opened a console: one column of plain text, a prompt, and slash commands. It worked, but you had to know what to type. Autopilot showed up as one-line change notices above the prompt, so the item waiting on you scrolled away with everything else. And in the October mockups I had picked a two-column layout, which the build quietly narrowed to one column.
So I went back to the mockups and picked again: a dashboard in the style of k9s or lazygit.
What you see now
On a terminal of at least 60 by 20, aidlc opens full screen. A header names the project, the branch, the autopilot run, the agent and what the session has spent. Under it sits an AUTOPILOT panel with one coloured row per item, then ROADMAP, INSTANCES and SPEND, side by side from 100 columns and stacked below that.
Everything is a key on what is selected. a approves an item waiting for review, with no confirm. r asks how many items and starts a run. p pauses or resumes it, and s stops it after a y. n names the next roadmap item and starts it in a worktree. ? lists the rest.
Ctrl-K replaces the slash. It is one list of every command, autopilot action, roadmap action and agent switch. Type to filter it. Words that match nothing become "Ask Claude Code: ..." and go to the agent.
The agent lives in a drawer, opened with c, under the autopilot panel, which stays on screen. /doctor and !git status still work in there, for people who already type them.
Anything running spins. Anything waiting on you stands still, in red.
Why this way
Every key runs the aidlc command it names, in the same process, with anything you typed passed after --. Approving from the dashboard runs the same approve command you would type, with the same output and the same refusals. I did not want a second path for each action that could drift from the first.
Commands that ask questions, like aidlc start, now ask them in a small form drawn over the dashboard. That took one change in the prompt module: a prompter scoped to a single call, so I did not have to thread one through every command.
What it does not do
There is no mouse, and you cannot select text inside the full screen. When you leave, the conversation is printed to your terminal, or saved to a file only you can read once it passes 200 lines. Below 60 columns you still get the old one-column console. And q never stops a run: it tells you autopilot is still going and how to watch it.