Skip to content
Documentation

AI agent slash commands and key bindings

Type / in the agent to open a filterable list of every command, including your custom commands and skills. /help prints the same list with key bindings. Text that starts with / but is not a command, such as a path like /etc/hosts, is sent to the model as a normal message.

Commands for features your plan does not include are shown greyed out.

CommandDescription
/helpCommands, key bindings, and a link to these docs.
/newStart a fresh conversation in this directory.
/clearStart a fresh conversation and clear the screen.
/resume [id]Reopen any conversation from any surface.
/continueReopen the latest conversation from this directory.
/rename <title>Rename the conversation everywhere it appears.
/rewindRestore files to their state before an earlier message.
/export [file.md|file.json]Export the transcript as Markdown or JSON.
/copyCopy the last answer to the clipboard.
/exit or /quitLeave the agent. Asks first if background tasks are running.
CommandDescription
/model [name]Show the current model with a menu, or switch. /model auto, /model register, and /model remove <name> are also available.
/models [name]Pick a model: Auto, platform models, or models you registered with your own key.
/effort low|medium|high|max|offReasoning effort for models that support it.
/contextHow much of the model’s context the conversation uses, with a breakdown.
/compact [focus]Compact the conversation now, optionally saying what to keep.
/usage or /costAI credits used in this session and this billing period.

See Models.

CommandDescription
/plan [off]Enter plan mode: read-only research that ends with a plan you approve.
/todosShow the agent’s current task list.
/agentsList agents and running subagents; run, create, edit, or delete one.
/tasksList background subagents, shells, and loops; tail, attach, or stop one.
/shells or /procsList running background shells and processes with PIDs, ports, and outputs.
/loop [--max N] [interval] <prompt>Run a prompt repeatedly; /loop stop [id] ends it.
/worktree [name]Move the session into a new git worktree; /worktree exit returns.
/review [branch | !<mr>]Review your pending changes, a branch, or a CHAMPREP Code merge request.
/security-review [branch]Security-focused review of the pending changes or a branch.
/diffEvery file change made in this session.
/undoRevert the last set of changes.

See Subagents, background tasks, and loops.

CommandDescription
/mcpMCP servers window: hosted and local servers, status, tools, and actions. See MCP servers.
/add-dir <path>Let the agent use another directory for this session.
/permissionsShow and edit allow and deny rules; mode switches the permission mode.
/servicesWhich CHAMPREP services the AI may use.
/skills [name]List skills. Run one with /<skill-name>.
/commandsList custom commands; /commands new <name> creates one.
/hooksShow, add, enable, and disable hooks.
/memoryView or edit CHAMPREP.md files and see your account memory.
/initSurvey the repository and write a CHAMPREP.md.
/trust [approve|revoke]Show or change trust for this repository’s .champrep/ configuration.

See Customization and Permissions and safety.

CommandDescription
/loginSign in with the browser.
/logoutSign out of this environment.
/profileSwitch the acting profile: personal, team, or organization.
/statusVersion, profile, environment, model, permission mode, and local paths.
/doctorCheck installation, sign-in, connectivity, plan, local tools, and terminal.
/feedback <message> or /bugSend feedback, optionally with a redacted excerpt.
/upgradeShow which version you run and how to update.
CommandDescription
/logoPlay the animated glowing ChamPrep startup logo in brand blue and cyan.
/color [name|reset]Accent colour for this project’s sessions.
/theme [auto|dark|light|high-contrast|colorblind]Colour theme.
/statuslineChoose status line segments or use your own command. Includes shells, agents, and tasks.
/output-style [default|concise|explanatory|learning]How answers are written.
/vimToggle vim keys in the prompt editor.
/keybindings [edit]View or change key bindings.
/configOne settings screen for all of the above.

See Customization.

KeyAction
EnterSend the prompt.
Alt+Enter or Ctrl+JInsert a newline. Ending a line with \ also works.
EscInterrupt the running turn, including running tools.
Esc EscOpen /rewind.
Shift+TabCycle the permission mode: ask, accept-edits, plan.
Ctrl+OExpand or collapse tool output.
Ctrl+TShow or hide the task list.
Ctrl+RSearch prompt history.
Up / DownPrevious or next prompt.
TabComplete a /command or @path.
Ctrl+LRedraw the screen.
Ctrl+CClear the input. Press twice to exit.

Change bindings with /keybindings edit, which opens $CONFIG/ai/keybindings.json.

SyntaxEffect
@pathAttach a file from the workspace to the prompt. Tab completes paths.
@agent-name promptSend the prompt to one of your agents as a background subagent.
!commandRun a shell command yourself in the workspace. Its output (with secrets redacted) is added to your next prompt.
#textSave text to the project memory file, CHAMPREP.md.

!command is run by you, so it does not show a permission prompt. It still uses a timeout and never receives your CHAMPREP credentials.