Documentation

Issue Runner — Batch AI Issue Fixing

Run AI against multiple open issues sequentially, either from the Issues page or as a standalone window.

Turns an issue straight into an AI-driven fix — pick one from the list and have the model work it in a real terminal, with live output and a git panel to review the changes as they land.

Issue Runner — issue list on the left, live terminal output on the right
Issue Runner — issue list on the left, live terminal output on the right

Layout

A two-column layout: the issue list (filterable, defaulting to Open & In Progress) on the left, and a working xterm.js terminal plus a Git Changes panel on the right. Each list row's icon reflects live run state — spinner while running, checkmark on success, ✕ on failure.

Running

  • Run Selected — runs the chosen issue interactively, so you can respond in the terminal if the CLI tool asks something.
  • Run All — runs every open/in-progress issue non-interactively, automatically chaining to the next runnable issue as soon as one succeeds. The chain stops on the first failure or when nothing is left to run.
  • Skip Permissions toggle (on by default) controls whether the CLI tool asks before each file edit/command — Run All always runs non-interactively regardless of this toggle's setting for a single run.

Typing exit in the terminal and pressing Enter closes the runner. Re-selecting a completed or failed issue replays its recorded output rather than showing whatever is currently running.

Git panel

Same file list / diff viewer / commit bar as elsewhere, pre-filling the commit message as Fix #<id> - <title>.

ShortcutAction
Ctrl/Cmd + EnterRun the selected issue
EscapeStop the current run
When the agent finishes without an error, the issue's status is automatically set to Resolved. This doesn't mean tests or success criteria passed — review the issue's git diff to confirm the fix.