Watch a specific run
Point it at any run id and drop straight into the cockpit.
Full-screen TUI monitor
A detached run used to be a black box you poked at with JSON. The monitor gives you a live cockpit: logs, a timeline, and approval gates in place, plus the power to grab the wheel when an agent gets stuck.
Why this exists
You kick off a long job in the background and then fly blind: polling status commands, grepping JSON, guessing whether it is stuck or just slow, and unable to step in when an agent goes down the wrong path.
What it shows
Streamed logs as steps run, so you see what the agent is doing the moment it does it.
Every step and its state in one column. Done, running, queued, failed, at a glance.
When a run pauses for a human answer, approve or reject it right here. It detects the gate for you.
An agent stuck in a loop? Drop into hijack mode and steer it back on course without killing the run.
Scroll back through the run's events, even for a run that started before you opened the monitor.
Built on OpenTUI. Move, toggle, approve, and switch modes without ever reaching for a mouse.
How to open it
Point it at any run id and drop straight into the cockpit.
Start a workflow interactively and the monitor opens on it as it runs.
Open the cockpit on any run and watch it, approve it, or steer it, all from the terminal.