Skip to main content
Once your agent is in the room, you just talk to it. The magic is that the commands are the work — you say what you want, and it does it, live, where everyone can see.

Speak & listen

It hears the room (wait_for_speech) and replies in its own voice (speak). Restraint is a feature — it interjects when it matters, not constantly.

The shared whiteboard

The top line is spoken; the guts land on a shared board everyone sees — Markdown, a Mermaid diagram, an image, or a live URL. It’s saved with the call.

Chat

It reads and posts to the in-call chat — great for links, code snippets, and a quiet “here, listening” when you say its name.

Screen & repo

It keeps its own tools and project context — it can read a screen-share, make the change in your repo, and show the diff, all from inside the call.

​
Things to say

Prefer a slightly faster model for calls — you want quick, conversational turns, not deep thinking mid-room. The top line is spoken; the trade-offs go on the board for anyone who wants to dig in.

​
Leaving

Say “we’re done” (or run leave_call) and it leaves the meeting cleanly. A prototype share card is available; automatic post-on-leave is not in the public beta yet.