The tools, by name - Vibeconferencing

Humans: you never type these — you just talk. This page is the tool reference for the agent driving your bot.

Listening and speaking

Tool	Use it for
`wait_for_speech`	Block until a human finishes a turn. This is the loop — call it again after every reply.
`speak`	Say something out loud. Pass `urgency` honestly; it gates whether you may talk over someone.
`set_mode`	`active` (reply freely) · `passive` (speaks only when named) · `silent` (never speaks). ⚠️ Passive AND silent both name-gate `wait_for_speech` — unnamed conversation is dropped before you see it (`local-server.js:2245-2272`). Silent is not “hear everything, say nothing.”
`send_chat` / `read_chat`	Post or read Meet chat — good for links and anything awkward to say aloud

Showing things

Tool	Use it for
`share_tab`	Share a live browser tab. The tab must be in a different Chrome window than the call.
`update_whiteboard`	Put markdown, Mermaid, an image, or a URL on the board
`start_share` / `stop_sharing`	Start and stop presenting. `start_share` is the primary — it takes `share_type`, `title_bar`, and a size. (`share_whiteboard` still exists as an older alias; prefer `start_share`.)
`click_share` / `type_share` / `inspect_dom`	Drive what’s on the board — operate a web app live in front of the room. Prefer selectors found with `inspect_dom`; `click_share` also accepts `x`/`y` for canvases, maps, and other content selectors can’t reach (`server.js:1844-1907`).
`set_share_title_bar` / `set_share_size` / `set_share_audio`	Shape of the shared window, and whether the room hears its audio
`scroll_share`	Scroll whatever you’re currently presenting
`get_call_screenshot`	See your own Meet view — including someone else’s share. (Meet has no canonical room view; layouts differ per participant.)
`get_shared_screenshot`	See your own whiteboard (Meet never shows you your own presentation)

Identity and state

Tool	Use it for
`set_voice`	Change voice mid-call. Accepts exact OS voice names, Voicebox names/IDs, and ElevenLabs names/IDs (`server.js:939-1071`) — use `list_voices` for what’s on this machine.
`set_avatar_emoji`	Change your face
`get_room_info`	Participants, share state, whiteboard content, your own call status
`list_preferences` / `set_preference`	Read and write your tunable settings

Leaving

Tool	Use it for
`leave_call`	Hang up
`end_session`	End the after-call work phase when you’re finished with it

The board is ROOM state, not yours

One whiteboard per room, shared by every bot. update_whiteboard writes room state and start_share presents that state (main.js:1073-1103, :3481-3483) — so any bot’s update changes what the room sees, and updating before you share is fine; nothing overwrites it.
Corollary: if another bot is presenting, don’t start_share — just update_whiteboard and your content appears on the board that’s already up. Re-sharing grabs the presentation.

An earlier revision of this page said the opposite on both points (per-bot boards, update only after sharing). Both were wrong against source — corrected Aug 9.

Shared whiteboard access requires signing in from App Settings. Screen Recording permission is optional and is used when sharing the board.