What shipped.

The insights dialog lands. Per sprint: tracked vs estimated vs logged, with a per-ticket breakdown and overruns called out plainly.

• ✳Insights dialog: tracked / estimated / logged per ticket, per sprint
• ✳Overrun flags — FD-123 ran 2.3× its estimate right on the board
• ✳Idle-gap sessionization tuned: default gap is now 4 minutes
• ✳Time values everywhere render in mono, because they're numbers, not vibes

One click on a Jira sprint now scaffolds a worktree per ticket — named, branched, and linked back to the ticket.

• ✳Sprint import: tickets → worktrees, with branch naming you can template
• ✳Ticket hover cards show description, assignee, and tracked time
• ✳Board transitions now cover custom Jira workflows, not just the defaults
• ✳Engine memory trimmed under long agent sessions (58 MB steady state)

Sessions moved into the engine. Close the window, quit the app, restart the machine — your Claude Code, Codex, and shell sessions come back mid-thought.

• ✳Persistent PTY sessions owned by the local engine
• ✳Super-tabs: agents, terminals, and the embedded editor grouped per worktree
• ✳Embedded VS Code opens in the worktree's root, keybindings intact
• ✳Dev server management: start, stop, and port status per worktree