Slash commands
Type / in the prompt to see commands that match what you type; Enter runs the highlighted one.
Ctrl+P opens the same list as a palette, and /help does too.
Text after a command is its argument, for example /messages migration searches your messages
for "migration".
What if I mistype a command?
A /word that matches no command is not sent to the agent. The status line says
Unknown command and, when the word looks like a typo, suggests the closest command, such as
/status for /stauts. Your text stays in the prompt so you can fix it. The same goes for a
command whose arguments it cannot use: the line stays, with the reason.
To send a message that starts with / as ordinary text, start it with a space. A path such as
/usr/bin is missing from PATH is sent as usual.
Chats
| Command |
Does |
/new |
Start a fresh chat. The old one keeps running if busy |
/sessions |
Open another chat. → shows actions; --all includes archived |
/fork |
Copy this chat up to a message you choose, and continue the copy |
/rename <title> |
Rename this chat |
/archive |
Hide this chat and open a fresh one |
/worktree |
Switch project folder, or create, reset or remove a worktree |
/copy |
Copy the last reply |
/export |
Write the chat to a Markdown file in your temporary folder and copy its path |
/share |
Publish a read-only snapshot of this chat on your Haarness server for seven days, or revoke its link |
/remote-control |
Continue this chat from your phone or any browser; /remote-control stop ends it. More |
/exit |
Quit the TUI. Running work carries on |
/share shows you the text it will publish before it creates a link. Anyone who can reach your
Haarness server can open the link. When that server only listens on this computer, as it does
unless you serve it on the network, the link opens only on this
computer, and only while Haarness runs; use /remote-control to work on a chat with someone else.
Model and agent
| Command |
Does |
/connect [provider] |
Connect a provider |
/providers |
Manage connections and custom endpoints |
/models [provider/model] |
Switch model (F2) |
/agents [name] |
Switch agent (Alt+A) |
/plan, /build, /review, /explore |
Switch to that agent; any text after it is sent |
/reasoning [level] |
Set reasoning effort (Alt+R) |
/ultracode [on|off] |
Toggle ultracode for this chat |
/router |
Choose the Haarness Cloud router this computer uses |
/billing [open|topup|usage] |
Your Cloud plan and credits; the words open that page in your browser |
While the agent works
| Command |
Does |
/cancel |
Stop the running turn now |
/queue [text] |
See queued messages, or queue one |
/input-mode [steer|queue] |
What Enter does while a turn runs, for this project. toggle switches, reset goes back to the default |
/tasks |
Watch a child agent's live transcript (Ctrl+O) |
/background [child-id] |
Let a waiting child agent finish in the background |
/btw [question] |
Ask a side question in a side chat |
/approvals [review|auto|yolo] |
Choose this chat's approvals: Review, Auto or Yolo (Ctrl+Y steps through them) |
/compact |
Summarise older context now to free room |
Plan and repeat
| Command |
Does |
/goal <objective> |
Work towards an objective across turns. See Goals |
/workflow [name] |
Browse or start workflows |
/loop <every> <prompt> |
Send a prompt again that long after each turn ends, e.g. /loop 10m check CI (30s, 5m, 1h) |
/loop stop |
Stop the loop. Loops also end after seven days |
/schedules |
Manage scheduled runs |
Review and history
| Command |
Does |
/diff [branch|last] |
Review changes full screen |
/undo |
Stop the agent and roll back its last change; your prompt comes back |
/redo |
Put back what /undo removed |
/timeline |
Roll back to any earlier point in this chat |
/messages [search] |
Search and reuse messages (Alt+M) |
/terminal [new] |
Open your terminal tabs |
Usage and status
| Command |
Does |
/usage |
Tokens, cache, cost and context for this chat (Alt+U) |
/status |
Model, agent, turn state, language servers, formatters and Chrome |
/stats |
Usage across projects |
/insights |
Save an HTML report of the same; --open opens it |
/usage-statistics |
Share daily usage counts with Haarness Cloud or stop, and see what is shared |
/logs |
The log level on this computer. /logs export [path] saves the logs as a zip with known secrets removed; /logs detail [on|off] switches detailed logging, which turns itself off after 30 minutes |
| Command |
Does |
/mcp |
MCP servers and their tools |
/skills |
Discovered skills |
/marketplace |
Search the Marketplace, install items and manage installed ones (/marketplace <text> searches, /marketplace installed) |
/plugin |
List, search, install or remove plugins |
/hooks |
Runtime hooks |
/memory |
What the agent remembers about this project |
/tools |
Turn tools on or off for this folder |
/profiles |
Save and restore configuration profiles |
/servers |
Manage Haarness servers and their projects |
/chrome |
Use your Chrome: set it up, turn it on or off, and choose the browser for new agent tabs |
/stash |
Drafts you saved with Ctrl+S (Alt+S) |
/editor |
Write the draft in $EDITOR (Alt+E) |
/themes |
Colour theme |
/language |
Interface language |
/keybinds |
Rebind shortcuts and see the fixed keys |
/tools lists every tool the agent can use, with an estimate of how much context its description
costs. Enter turns the highlighted tool on or off for this folder, and the setting is kept.
| Form |
Does |
/tools disable <tool> |
Turn a tool off |
/tools enable <tool> |
Turn it back on |
/tools reset <tool> |
Back to your configuration |
/tools show, /tools hide |
Open or fold every tool row in the transcript |
/tools lazy auto|always|off |
When rarely used tools are loaded only on demand (reset goes back to your configuration) |
Your own commands and workflows
Commands you define, commands from plugins, and workflows you have added appear in the same list
and run the same way. If one has the same name as a built-in command, the built-in wins. See
Agents, commands and skills.
Other names
Many commands also answer to another name.
| Command |
Also |
/new |
/clear |
/sessions |
/resume, /continue, /threads |
/worktree |
/worktrees, /workspace, /workspaces, /project, /projects |
/remote-control |
/rc, /remote |
/models |
/model, /mo |
/connect |
/login |
/providers |
/provider |
/agents |
/agent |
/reasoning |
/think, /thinking |
/ultracode |
/uc |
/router |
/routers |
/billing |
/credits |
/cancel |
/stop |
/tasks |
/subagents, /children |
/background |
/bg |
/workflow |
/workflows, /wf |
/schedules |
/schedule |
/timeline |
/history, /checkpoints |
/messages |
/message, /browse |
/terminal |
/term |
/usage |
/context |
/usage-statistics |
/telemetry |
/logs |
/log |
/mcp |
/mcps |
/marketplace |
/market |
/plugin |
/plugins |
/tools |
/details |
/profiles |
/profile |
/servers |
/server |
/chrome |
/browser |
/editor |
/edit |
/themes |
/theme |
/language |
/lang, /locale |
/keybinds |
/shortcuts |
/help |
/commands, /? |
/exit |
/quit, /q |