I'm building Aldena, a workspace where you hire a team of AI agents instead of prompting one assistant.

Every project gets its own room with a dedicated server, its own integrations and credentials, and its own memory. You hire agents into it from a roster of real roles, then wire them into an org chart so work travels down manager lines and results come back up. You watch it happen in a top-down view of the room.

Every tool an agent can reach runs under an ask, allow, or deny policy. Agents stop mid-run to ask you questions in plain language, and code arrives as a pull request no agent is allowed to merge. Nothing crosses between rooms, so one client's work stays put.