Split panes

Split any pane and you get a second, fully independent terminal session beside it — its own shell, its own working directory, its own history and scrollback, its own agent if you start one.

Splitting

Both are gated on the prompt input not having focus, so ctrl+D at the shell still sends EOF as expected. You can also split from a block's right-click menu ("Split pane right" / "Split pane down") or the command palette. Splits nest freely — split a split to build any grid you like.

Focus

Click anywhere in an inactive pane to focus it. A small accent triangle in the top-right corner marks the active pane, and the tab title follows the active pane's working directory. If you find the unfocused panes distracting, turn on Dim inactive panes in Settings → Appearance.

Closing

• Hover a pane's top-right corner and click the ✕ (red on hover).
• Or press ctrl+W to close the active pane.

Closing a pane collapses the split — the surviving sibling takes over the space. On a tab's last pane, ctrl+W closes the tab. Closed panes shut down their PTY and any running agent cleanly.

Resizing

Every divider is draggable: grab the hairline between two panes (it highlights in the accent color, with a resize cursor) and pull. Ratios are clamped so no pane can be squeezed out of existence.

Independence, concretely

• cd in one pane never moves another — see Sessions.
• Each pane's chips, Node version override, and environment are its own.
• Agent conversations are per-pane; split panes host independent agents side by side.

Splits everywhere

Tab configs can declare whole split layouts declaratively — a two-pane agent-plus-server workbench, for example — so ctrl+T can open your grid pre-built. See Tab configs. Every pane also appears as its own row in the sidebar rail.