> ## Documentation Index > Fetch the complete documentation index at: https://www.rumus.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Workspace > Arrange multiple terminals, monitors, and notes on a zoomable canvas — and broadcast commands to several at once. A **workspace** is a special tab type that opens onto an **infinite canvas** — a 2D surface where you can place cards (terminals, monitors, markdown notes), arrange them freely, and save the layout for next time. Use workspaces for project-shaped work: one workspace per service, environment, or incident, with the terminals and monitors you need for that job already laid out. ## Cards Everything on the canvas is a **card**. Three types ship today: | Card | What it is | | ----------------- | --------------------------------------------------------------------------------------------------- | | **Terminal card** | A shell session — local or remote — exactly like a terminal tab, just placed on the canvas | | **Monitor card** | A live view of CPU, memory, disk, and GPU metrics for a host — see [Monitoring](/docs/remote/monitoring) | | **Markdown card** | A free-form note for runbooks, command snippets, or status notes | Add a card with the canvas toolbar's **New Terminal**, **New Markdown**, or **New Monitor** buttons, or by right-clicking empty canvas space. ## Working with the canvas Middle-drag, two-finger trackpad pan, or hold Space + drag. Pinch on a trackpad, Ctrl/Cmd + scroll, or use the **Zoom in / Zoom out / Reset zoom** controls. Drag the card header to move; drag the corner to resize. Cards snap to a grid. Cards have a z-order. Click a card to bring it to the front; the canvas remembers your stacking. ### Toolbar actions The canvas toolbar exposes a few helpers: * **Select All** — select every card on the canvas (useful for [broadcast](#multi-terminal-broadcast)). * **Fit to Screen** — zoom and pan so all cards are visible. * **Auto Arrange** — re-tile cards into a clean grid. * **Minimap** — toggle a small overview map in the corner for navigating large canvases. * **Zoom In / Zoom Out / Reset Zoom** — adjust the zoom level. ### Multi-select Multi-select cards by: * **Drag-select** — drag an empty area of the canvas to draw a selection box. * **Ctrl/Cmd + click** — toggle individual cards in the selection. The most useful thing to do with multi-select is [broadcast a command](#multi-terminal-broadcast) to several terminals at once. ## Saving and switching workspaces A workspace persists everything it contains: cards, positions, sizes, z-order, and the canvas viewport (pan / zoom). The next time you open it, it looks exactly as you left it. * Open the **workspace switcher** in the top bar to switch between workspaces or create a new one. * Each workspace has a **name**, **description**, **icon**, and **color** you can edit from the switcher or from **Settings → WorkSpace**. ## Multi-terminal broadcast When you need to run the same command across several servers — a fleet update, a parallel restart, a quick health check — Rumus can broadcast a single command to multiple terminal cards simultaneously. It's a workspace canvas feature, available the moment you have more than one terminal card selected. ### When you'd use it Typical cases: * Run `df -h` across every node in a cluster and eyeball them side by side. * Roll a single `systemctl restart` to a small fleet during a maintenance window. * Fan out a quick `tail -n 50 /var/log/...` to see logs from each replica together. ### How to broadcast Two ways: * **Drag-select** — drag an empty area of the canvas to draw a box around the cards you want. * **Ctrl/Cmd + click** — toggle individual cards into the selection. The canvas surfaces the count: *"{count} terminal(s) selected"*. The broadcast input bar appears with the placeholder *"Enter command to send to {count} terminals…"*. Type a command and press **Enter**. Smart autocomplete works here too — partially type a command and accept the suggestion the same way you would in a single terminal. The command runs in every selected terminal in parallel. Each card scrolls independently with its own output, so you can compare results side by side. ### What gets broadcast * The literal text you type, followed by a newline. * Smart autocomplete suggestions, when accepted. That's it. If a target terminal is in the middle of an interactive prompt, your broadcast input is sent to that prompt — same as if you'd typed it locally. ### Caveats * **No "abort all"** — once typed, each terminal runs the command independently. To stop a runaway, focus the offending terminal and send your usual interrupt (`Ctrl+C`). * **Pseudo-terminal differences matter.** A command that depends on environment variables or shell aliases will only behave the same way if every selected terminal has those set up. * **Broadcast targets are terminal cards on the same canvas.** It's not a cross-tab feature — plain terminal tabs aren't selectable. * **Approval flows still apply.** If a target terminal has the AI agent active and command approval gating is on, the broadcasted command is treated like any other input. ### Tips * For a fleet of identical hosts, group them in a workspace with a clean grid layout (use **Auto Arrange**) so output is easy to scan. * Combine with **Monitor cards** ([see Monitoring](/docs/remote/monitoring)) to watch CPU / memory react to whatever you just rolled out. * Pair with [Skills](/docs/ai/rules-skills) so the AI can suggest broadcast-safe commands before you fire them across a fleet. ## When *not* to use a workspace Plain terminal tabs are still the right choice for ad-hoc work — quick one-off shells, scratch sessions, or anything you don't need to come back to. Workspaces shine when: * You repeatedly need the same set of terminals and monitors together. * You want to keep prod / staging / dev environments visually separate. * You're running an incident response and want notes alongside live shells. ## Next steps Add a live CPU / memory / disk / GPU chart for any host. Save the SSH targets you broadcast to.