Skip to content

Slash Commands

Slash commands let you control your PizzaPi session directly from the chat input. Type / to open the command picker, then select or type a command.

Type / at the beginning of the chat input. A popover appears listing all available commands, grouped into sections:

  • Commands — built-in session and UI commands
  • Extensions — commands registered by Claude Code plugins
  • Prompts — prompt template shortcuts
  • Skills — skill-based commands from your configuration

The list filters as you type. For example, typing /mo narrows the list to /model.

KeyAction
/ Move highlight through the list
EnterSelect the highlighted command
EscapeClose the picker

On touch devices, tap a command to select it.

Some commands (like /mcp) have sub-commands. When you select one of these, the picker stays open and shows the available sub-options. For example, selecting /mcp then shows status, reload, disable, and enable.

If a sub-command requires an argument (like /mcp disable), selecting it fills the input with the command prefix and positions your cursor to type the argument.


CommandDescription
/newStart a new conversation. If linked child sessions are still active, a confirmation dialog appears first.
/resume [query]Resume a previous session. Opens a session picker showing recent sessions with name, date, and path. Type to fuzzy-search by session name, ID, path, or first message.
/stopAbort the current generation.
/restartRestart the underlying CLI process.
CommandDescription
/modelOpen the model selector dialog.
/cycle_modelSame as /model — opens the model selector.
/effortCycle through reasoning effort levels (low → medium → high).
/cycle_effortSame as /effort.
CommandDescription
/compact [instructions]Compact the conversation context. Optionally pass custom instructions to guide summarization, e.g. /compact focus on the auth module.
/planToggle plan mode. In plan mode, the agent explores and plans without making changes (read-only).
CommandDescription
/name <name>Set a display name for the current session, e.g. /name auth-refactor.
/copyCopy the last assistant message to your clipboard.
CommandDescription
/pluginsList all loaded Claude Code plugins, including their commands, hooks, skills, agents, and MCP servers.
/skillsList all available skills from your configuration and plugins.
/sandboxShow the current sandbox status — mode (none/basic/full), platform, and recent violations.

The /mcp command manages Model Context Protocol servers. Type /mcp (with a space) to see sub-commands:

Sub-commandDescription
/mcp statusShow the status of all configured MCP servers.
/mcp reloadReload MCP server configurations.
/mcp disable <name>Disable a specific MCP server by name.
/mcp enable <name>Re-enable a previously disabled MCP server.
CommandDescription
/agents [name]Start a new session as a named agent. Without an argument, opens a picker showing all available agent definitions. With a name, spawns directly, e.g. /agents reviewer.

The agent picker shows each agent’s name and description. Select one to start a new session with that agent’s system prompt.


In addition to the built-in commands above, three types of commands appear dynamically based on your configuration:

Skills defined in your ~/.pizzapi/skills/ directory (or registered by plugins) appear in the Skills group. They are prefixed with skill: internally but shown by their short name in the picker. Running a skill command sends its content to the agent.

Claude Code plugins can register custom commands. These appear in the Extensions group. Each shows the command name and description provided by the plugin.

Prompt templates from your configuration appear in the Prompts group. Selecting one inserts the template content into the conversation.


/new # Start a fresh session
/resume auth # Find and resume a session matching "auth"
/compact focus on API routes # Compact context with custom instructions
/name deploy-fix # Name the current session
/mcp status # Check MCP server health
/mcp disable filesystem # Disable the "filesystem" MCP server
/agents reviewer # Start a session as the "reviewer" agent
/effort # Cycle to the next reasoning effort level
/plan # Toggle read-only plan mode
/copy # Copy last response to clipboard