The AI is in chat, on tasks, in discussions, on pages and in huddles — with a name, an avatar, a brief and a profile card. It replies where the conversation is, reacts when a paragraph would be too much, and authors every change as itself so you always know what it did.
You go to it, you tell it everything, you copy the answer back. It never sees what happened while you were away, never speaks up when it knows something, and never leaves a trace anyone else on the team can follow.
The help exists, but only for whoever thinks to open the panel — and only for what they thought to paste.
If output lands as your message, nobody can tell which decisions were the AI’s and which were yours.
A research question and a code review want different behaviour, and you get exactly one.
Every workspace starts with an AI teammate — name, avatar, profile card and a brief an admin can edit.
Chat, a task comment, a discussion, a page, a huddle. It answers where the conversation already is, and it can act while it’s there.
A researcher, a QA reviewer, a planner — each with its own brief, model, tools and behaviour. Mention the one you want.
Length, tone, how hard it pushes back, whether it thinks first, whether it cites sources, which language it answers in — all set per agent by an admin.
An agent with no create-or-edit tools looks things up and answers, and says so plainly when asked to change something.
A classifier decides reply, react with one emoji, or nothing — and “nothing” is the common case. Ordinary teammate talk is left alone.
Tasks, comments, pages and edits are authored as the agent, so the activity feed shows exactly what it touched.
Free for up to 10 people, with 1,000 AI credits included. No credit card, no sales call, no onboarding deck.