Slash commands
Type / in an agent or terminal pane to run hob commands and your backend's own commands.
Slash commands are shortcuts you type in the input box of an agent or terminal pane.
Start typing / and a menu appears, filtered as you type — /ag narrows to /agent.
Using slash commands
Type / in an agent or terminal pane
The slash menu appears above the input. (Render, web, and selection panes don't have one.)
Pick a command
Use the arrow keys to move through the list, or click a row directly. The menu filters as you keep typing.
Run it or fill it
Press Enter to run the highlighted command. Press Tab to insert it into the input
without running it — handy for backend commands that take arguments.
Enter runs the command; Tab only inserts it so you can add arguments first. Commands
that expect input (skills, custom commands, MCP prompts) insert themselves with a trailing
space, ready for you to type.
hob commands
These are built into hob and work the same across backends:
| Command | What it does | Where |
|---|---|---|
/agent | Replace this pane with a fresh agent — same backend, model, and permission mode, empty conversation | Agent + terminal panes |
/terminal | Replace this pane with a fresh terminal (inherits the working directory) | Agent + terminal panes |
/exit | Close this pane | Agent + terminal panes |
/plan | Toggle plan mode | Agent panes (plan-capable backends) |
/session | Copy the backend session ID | Agent panes (after the conversation starts) |
/agent and /terminal replace the current pane in place at the same grid position —
they don't add a new pane alongside. /plan only appears for backends that support plan
mode, and /session only appears once a session exists.
Backend commands
In an agent pane, the menu has a second group below hob's commands, labeled with the active backend (Claude Code, Codex, OpenCode, or a custom backend). It surfaces that backend's own catalog:
- Built-in commands the CLI ships with
- Skills and custom commands you've defined
- MCP prompts from connected servers
Each carries a small source badge (skill, custom, mcp). Selecting one that takes an argument inserts it into the input so you can finish the prompt before sending; plain commands run immediately. Terminal panes show only the hob commands above.
Plan mode
/plan puts the agent in plan mode (the same toggle as Alt+P). Instead of changing
files immediately, the agent drafts a structured plan first. You review it in a collapsible
panel, then approve — choosing whether the agent asks before each step or executes the plan
autonomously.
Plan mode works well when a task spans multiple files, needs research, or you just want to
review the approach before any code is written. See
Parallel agents for permission modes, and
Keyboard shortcuts for the Alt+A / Alt+T ways to
create panes without the menu.