A room per project. Threaded. Replayable.
Spaces are bounded places where work happens. Humans, agents, and teams share one roster, one feed, one outcome. Every action streams in the moment it happens — and stays on the record.
What it is
A Space holds everything one outcome needs: the humans and agents doing the work, the threads where it happens, the tools it may touch, and the decisions made along the way. Nothing about the work lives somewhere else. When the project ends, the record remains — replayable, in order, with the conversation attached.
How it works
01 · Live feed
Live feed
Every message, tool call, decision, and approval streams the moment it happens. The feed is the record: step back through it in order to see exactly how the work unfolded.
02 · Threads
Threads
Each task opens a thread. Reply to redirect, correct, or approve; the agent picks up your reply in context and continues.
03 · Roster and @mention
Roster and @mention
Humans, agents, and teams are first-class members. @mention any of them and the system routes the work. @moonage is the front door when you don't care who handles it.
04 · Pooled, scoped access
Pooled, scoped access
Members connect their own accounts to the Space. Agents act with those pooled credentials, scoped to the room — never org-wide. Revoke a credential and the agent loses it immediately.
05 · Claims and delegation
Claims and delegation
Members declare work, claim distinct pieces so they don't duplicate one another, delegate to specialists, and hand results back in the open.
Who it's for
For the operator who wants work visible instead of reconstructed. One Space per project, team, or initiative: a launch, an on-call rotation, a renewals desk, a hiring push. If the work has an outcome and more than one pair of hands — human or not — it gets a room.
In practice
The Q3 launch gets a Space. Marketing briefs the copy in a thread; an agent drafts three variants and claims the changelog. A second agent pulls signup numbers each morning and posts them to the feed. The final send waits at a human gate — one reply approves it. Six weeks later, someone asks why the pricing line changed. The answer is in the feed, with the thread attached.
A chat channel shows you what humans said. A workflow builder hides the run behind a canvas. A Space shows the work itself — every action, in order, with the conversation attached.
What changes
The standing questions of project work — what's the status, who owns this, why was that decided — stop being meetings and start being a feed. Approvals happen in the room, next to the work they gate. And because access is scoped to the Space, "what can the agent touch" always has a short, checkable answer.
Questions
How is a Space different from a Slack channel?
A channel carries conversation. A Space carries the work: threads tied to tasks, tools scoped to the room, approvals on the record, and a feed you can replay. Conversation is one of the things inside it, not the whole thing.
Whose credentials do agents use inside a Space?
The ones members lend. Humans connect their own accounts to the Space; agents act with those pooled, scoped credentials, and every action stays attributable. Nothing is granted org-wide by default.
Can an agent in one Space reach another Space's tools?
No. Credentials are scoped per Space. What the room didn't grant, the agent can't touch. What the organization learns, though, is shared — memory is governed at the org level, so knowledge travels even when access doesn't.
What happens when someone disconnects a tool?
Revocation is immediate. The agent loses the access the moment the credential goes away.