Slash commands#
Type these in the chat box. /help lists them. Anything else that looks like /word is reported as unknown
instead of being sent to her, while paths like /etc/hosts are sent as usual.
| Command | What it does |
|---|---|
/help | list the commands |
/welcome | a short tour of what is here and the command for each thing; it is also shown once on the first start of a new data folder |
/compact [focus] | summarise the conversation to free up context; she keeps the gist |
/clear, /new | start a fresh conversation, like New chat |
/resume [number] | list earlier conversations in the current folder (newest first), or continue one |
/cost | what the conversation has used: an API-price estimate and how full the context is |
/model [name] | show or set the model for the current mode, until the server restarts |
/chat, /work [folder] | switch modes |
/permissions [ask|plan|accept|auto|everything], /plan | the permissions menu, from the keyboard |
/effort [level] | show or set how hard she thinks (restarts her process; the conversation continues) |
/setup [on|off] | work mode: load your own Claude Code settings, hooks and MCP servers (off at every start) |
/chrome [on|off] | work mode: let her use your Chrome through the Claude in Chrome extension (off at every start; the extension must already be installed and signed into the same account) |
/persona [show | edit | reset] | whose voice she speaks in (your own persona, the character of the pack that is on, or the default); edit opens your own data/persona.md, reset removes it (kept as persona.md.bak); both only on the computer |
/memory [add <note> | forget <number|all> | edit | ask | auto | off] | her notes about you (data/memory.md); in chat she asks before saving one (auto saves with an Undo, off saves nothing), never in work mode |
/card import <file> | apply | cancel | export | reset | character cards: use a card's personality, or export hers (/card import shows a preview first) |
/skills [on|off] | work mode: let her use your own skills by name; your user settings apply, hooks stay off (off at every start) |
/studio [on | off | handover [outfit] | backend <address> | budget <dollars>] | make a picture pack of your own with an image model (the pack studio, a draft maker; off at every start, only on the computer). backend sets the model's address (plain http on this computer, or grok-web / gemini-web / chatgpt-web for the Grok Imagine, Google Gemini or ChatGPT website driven through Claude in Chrome (it uses Claude Code and your Chrome), or xai, gemini or openai for a paid service, which needs XAI_API_KEY, GEMINI_API_KEY or OPENAI_API_KEY and counts every picture against budget); on checks it and opens the panel (even when the maker is not ready: it says what is missing, and the panel's list can choose another); handover prints a prompt to copy into any AI agent that makes pictures (it works with the studio off) |
/rename [name | reset] | give her a name of your own (it wins over the name of a pack's character); reset, or the reset beside her name, goes back to the character's name, or Risa |
/frames [name|off] | show her with a picture pack; alone, lists the packs |
/play [name] [end] | play a short scene in chat (/play lists them; /play key is The spare key; /play room is a short story you are both trapped in, and the page asks you to wait while she writes it; /play room 5 1 3 1 sets steps, hints, herrings and traps; a trap starts that room over; she explains the thread when you finish or /play end). In work mode, /play switches to chat |
/show | she plays every emotion, a blink, talking and her gestures, naming each |
/phone [devices|revoke <n|all>|new|off] | a link and QR code that let one phone connect (you allow it on the computer); with phone access off it asks whether to turn it on (until the next start); devices lists the phones you allowed, revoke removes them, new makes a new link, off turns it off |
/status | brain, model, mode, folder and session id |
/restart | restart the server. The page reconnects, and the conversation stays. Phone access and permissions go back to how they are at every start. Not from a phone |
/doctor | check what the app needs (Node, the character creator, the brains, which opencode it uses, her data folder) and say what is missing and how to fix it; it only looks, changes nothing and is not available from a phone. Paste its output into a bug report |
/claude-login | sign in to Claude Code again: it runs Claude Code's own claude auth login, which opens your browser (the app never sees your credentials), then checks claude auth status. When her login has expired the error says to use it; not available from a phone |
/install-opencode | install the opencode this app was made for (a fixed version, checked against a known checksum) into the app's own folder, for the app only: it asks first, is used instead of any other opencode on the computer, and leaves yours untouched |
/copy | copy her last reply to the clipboard |
/export | download the conversation as a Markdown file |
/voice | turn spoken replies on or off |
/rc, /remote-control | not built in: explains why and prints the terminal command to use it yourself |
/rewind | not available in the companion: prints the terminal command to continue there and rewind |
Good to know#
/resumelooks in the folder she is working in (the project folder in work mode, her ownworkspace/in chat mode), so it also finds sessions you started in a terminal there. It is refused mid-reply and with Ollama.- Commands that change settings are refused mid-reply.
- Commands that need Claude Code (
/compact,/cost,/resume,/work,/permissions,/plan,/rc,/rewind) say so plainly when another brain is active, instead of pretending to work. - Some commands are refused from a phone: see From your phone.
Made with AI coding agents, mostly Claude Code. This documentation was written by Claude Code. Details