Your whole team, in one live room.

Solo Jarvis is a conversation with one supervisor. The War Room generalises it to many: the supervisor moderates as the amber “Vanta”, and every live agent joins the roster as a voice you can address. The shape is hub-and-spoke — you and the supervisor are the conversational core, and agents answer when asked.

Address an agent by leading with its name (“scout, …”) or tap its chip in the roster. Your next turn routes there until you switch — and the central orb tints to whoever is holding the floor.

Each participant is assigned a distinct local neural voice, handed out so the supervisor and the first agents always sound different. A single serialized speech queue means replies never talk over one another — each line plays in its speaker’s voice, one at a time.

• The supervisor speaks in amber as “Vanta”; each agent carries its own identity colour.
• One serialized queue — no overlapping speech, and replayed history is never read aloud.
• The roster shows who’s live, who’s speaking, and where your next turn will land.
• Push-to-talk or open-mic — the same on-device engine as Vanta Voice and Jarvis.

When the room reaches a decision, the supervisor — the only participant with memory tools — writes a concise, structured note to the project knowledge graph. Each one surfaces live on the Decisions rail as it lands, newest first.

Worried something slipped by? Say “log that decision” (or tap Log decision) and the supervisor captures the recent exchange deterministically — so the record forms as you talk, not after.