A stream is a shared conversation for a team, a topic, a project or the whole company. AI agents are members too. They answer in the thread, brief anyone who was away and track what was promised, so both the conversation and the work move faster.





Mention an agent by name and it replies in the thread. It can read the stream's history, so it answers from what the team has already decided.
Ask an agent something in a stream and watch it work. It shows what it is doing now, how long it has been going and what it has finished, so nobody is left waiting without knowing why.
Press Catch me up and the system agent writes a short brief of what was decided, what is stuck and what is waiting for a reply.
When someone asks for something or agrees to do it in a stream, it is picked up as a to-do with an owner and a due date. Nobody has to write it down.
In a broadcast stream only admins post, and replies come back as whispers. A whisper is a private side conversation attached to a message. It works in any stream, for a remark that is not for everyone.
One flowing chat with threads. Good for a team that talks all day.
Every post starts a titled topic, like a forum. Good for large groups and decisions.
Admins post and everyone reads. Replies arrive as private whispers. Good for company news.
Messages disappear after 24 hours. They are never searched, archived or remembered by agents. Good for casual talk.
Anyone in the workspace can join a public stream. Private and invite-only streams are for groups that need a closed conversation, such as hiring or security. An archived stream stays readable after the work is done.
Chats are conversations with customers, in the omnichannel inbox. Streams are the conversations inside the company.
Yes, if it has permission. A digest agent opens one every morning, and anything it proposes still waits for a person.
Only for the people in them. For everyone else a whisper does not appear in search or in the stream.
What its role allows. Each stream lists the agents in it and what they may do there, for example read everything or only write drafts.
Start with one stream, add an agent, and see how much sooner a question gets answered.