ncode CLI Using ncode
Slash commands
Type / in the composer to list every command; the list narrows as you type, and Tab or Enter takes the highlighted one. A command that needs no argument runs at once. Commands work while other runs are working.
Conversation#
| Command | What it does |
|---|---|
/new (also /clear) | start a new conversation; this one stays saved |
/resume | pick a saved conversation by title |
/resume <id or title> | open that conversation |
/resume-run | resume the last stopped run of this conversation |
/stop | stop everything running in this conversation |
/queue <text> | queue a message behind the running turn |
/compact [focus] | summarise the conversation so far and continue from the summary |
/goal <text> | set the conversation goal every agent keeps in mind; alone, show it |
/search <words> | search the messages of your conversations |
/export [file] | write this conversation as Markdown |
/cost | tokens and cost of this conversation, by model |
/attach <image-path> | stage an image file from the project for the next message |
/deep_research [id] | attach a finished deep research to the next message |
Runs and modes#
| Command | What it does |
|---|---|
/swarm <task> | start a swarm of agents on a task |
/plan [task] | turn plan mode on or off; with a task, plan it now as a read-only run |
/consensus [task] | switch to Consensus mode; with a task, run this turn as a judged plan |
/review | review the uncommitted changes and report problems |
/ultra | turn Ultra on or off: big tasks become workflows |
/workflow <name> [key=value…] | launch a workflow |
/workflow pause|resume|stop|save <run> | control a workflow run |
/workflows | open the workflow library and runs |
/create-workflow [what it should do] | write a new workflow with the assistant |
/rewind | restore files to how they were before an earlier turn |
/agents | the agent definitions a swarm can use |
See Modes and run types and Workflows.
Model and effort#
| Command | What it does |
|---|---|
/model [model] | switch this conversation's chat model; alone, a picker |
/swarm_model [model] | switch the model the sub-agents use |
/effort <level> | reasoning effort of the chat model |
/swarm_effort <level> | reasoning effort of the sub-agent model |
Approvals#
| Command | What it does |
|---|---|
/approval [read-only|auto|full] | set the project's approval mode; alone, a picker |
/trust | trust this project: read its instructions and allow edits |
Screen and settings#
| Command | What it does |
|---|---|
/panel full|compact|hidden | size of the side panel |
/panel summaries on|off | AI status lines in the side panel |
/diff on|off | tool rows with or without their diffs and previews |
/theme [dark|light] | switch the colour theme at once |
/mouse [on|off] | mouse-wheel scrolling on or off |
/settings [what] (also /config, /prefs) | open Settings, at a section or setting |
/help | the commands and the keys |
/quit (also /exit) | leave; running work of this session stops |
/panel, /diff, /theme and /mouse are remembered for the next session. To see which files this conversation changed, open Changes from the Ctrl+P palette (see Rewind and changes).
Your own commands#
Custom commands and saved workflows appear in the same / list. A Markdown file in the project's .swarm_code/commands/ folder becomes a command of its own; see Project instructions, memory and config. If a name is taken twice, a built-in command wins over a workflow, and a workflow over a custom command.
Note On the desktop app /resume resumes the last stopped run; in the terminal that is /resume-run, and /resume opens a conversation.