Runs

A Run is one execution of a workflow. Its state belongs to Aqorin, not to any runtime session, so a runtime that crashes does not erase what happened.

Runs are detached

A Run runs in its own worker. Quitting the TUI, or closing the terminal, leaves it running. Run aqorin again, or aqorin logs <run> --follow, to follow it.

There is one Run at a time per working tree. A new Run or chat is refused while another is unfinished.

The Runs screen (2)

Every Run and chat of this project is listed with its state and when it started. An open chat reads chat open. / filters the list, n starts a new task, and c starts a Free chat.

A Run’s view

A Run’s view is a dashboard:

• On the left, the steps: each one’s state, its time, which agent runs it and where (agent · runtime), and its attempt.
• In the middle, the selected step’s work. One line says what it is doing now — Thinking…, Running tool… npm test, Waiting for you — and the next who runs it, with the model the runtime reports and, on Claude Code, the context it has used (of its window, once the step ends). Then come its messages. The tool calls between two messages are folded into one muted line, with their count and any that failed: t shows each one.
• On the right, the Run in a few lines: its state, workflow and assurance, task, time, changes and handoffs. An advisory Run says there, once, what that means. d shows why, in a few words, with the selected step’s details (its assignment, version and attempts); the full reasons are in its Result.

After a step whose work is checked, the steps show checking until the checks report: Aqorin runs the task’s checks — after the step’s runtime checks the work (advisory), when the task asked for the agent’s check. With no check and no agent’s check, nothing runs: the step moves on and its checks read not checked. The next step waits for them, so a reviewer that has not started yet is usually waiting on these checks. Meanwhile the step’s work shows what the checks are doing — the runtime’s messages and tool calls, then each of the task’s checks as it runs — and, once they report, keeps them under Checks.

Tab to the steps, then ↑ ↓ picks one. y sends the selected step’s messages to your terminal’s clipboard, where the terminal allows it (it cannot confirm it): a mouse selection takes whole lines across every pane. c c cancels the Run. Ctrl+T opens the Trace.

The result

A finished Run opens on its Result, which shows:

• how it ended, and why;
• the review;
• the checks;
• the changes;
• its assurance. A Run is advisory when its workflow says it is, has no step that runs checks or no review step, when the task had no checks, or when a read-only or review step ran with an advisory restriction. Otherwise it is enforced. An advisory success reads “completed — advisory”: it is never shown as validated or independently approved. The Result claims the assurance was reached only for a Run that succeeded.

From the CLI

aqorin runs
aqorin status <run>
aqorin inspect <run>
aqorin logs <run> --follow [--json-events]
aqorin cancel <run>
aqorin resume <run>

Ctrl+C on a CLI Run interrupts it at its next step boundary, and it can be resumed. A second Ctrl+C stops waiting, but the interruption still happens.