Skip to content

Workbench layout

The centre of the window is a workbench: a tree of tab groups separated by resizable splitters. You manipulate it exactly as you would a code editor’s.

A split workbench: the request's configuration on the left, its response on the right, and the console in a full-width group below

Drag a tab and drop it:

  • on another tab or in the centre of a group → it joins that group;
  • on an edge (left, right, top, bottom) → it creates a split.

A vertical split places the groups side by side; a horizontal split stacks them. Every splitter can be dragged with the mouse to readjust the proportions.

A tab can also be detached into a new window.

Drag a tab within its own group to change its position.

Double-click a tab header: it goes full screen over everything else. Esc brings it back. Only one tab at a time can be maximised.

Ctrl+Tab and Ctrl+Shift+Tab open a picker ordered by most recent use: releasing Ctrl confirms, Esc cancels.

Closing the active tab activates the most recently used one, across all groups — and moves the focus to its group if need be.

A tab that is moved, reordered or maximised keeps its instance: an HTTP response still loading, an unsaved edit, your scroll position or the active sub-tab are never lost. That is the opposite of the usual behaviour of a web interface, and it is deliberate.

The tab tree is saved per project. Reopening a project restores the same tabs, the same splits, and even the mini browser’s URL.

Five surfaces open as a single instance, from the View menu:

TabMenu
ConsoleView ▸ Console
FirewallView ▸ Firewall
Mini browserView ▸ Mini browser
SettingsView ▸ Settings
MCP logsView ▸ MCP logs

Opening the Console for the first time automatically places it in a new group below all the others, spanning the full width.

Three tools (Pro edition) let an agent read and rebuild the layout: get_workbench_layout, set_workbench_layout and set_tab_maximized. See MCP tools.