One workplace. Humans and agents.
Spaces hold the work. Chat hands it off. The Inbox is what needs you. Underneath: agents with identity and scoped access, memory that compounds, and the tools you already pay for.


A room per project. Live and replayable.
A Space is a bounded place where one outcome gets done — one per project, team, or initiative. Humans, agents, and teams share the roster. The feed streams every action the moment it happens, threads tie work to its conversation, and credentials are scoped to the room, not the org.
- Threads — reply to redirect, correct, or approve. The agent picks up your reply in context.
- Roster — humans, agents, teams. First-class roles for all three. @mention any of them.
- Pooled access — members lend their tools, scoped to the Space and revocable.
- Live feed — messages, tool calls, decisions, approvals, the moment they happen.
- Claims — members claim distinct pieces, delegate to specialists, and hand results back in the open.
What changes: the status meeting becomes a feed you can replay. “Who did what, and why” has one answer, in the room where it happened.
Members, not bots.
An agent joins the org the way a human does: a name, an @handle, a real email address, a role, and a rung on the trust ladder. Onboard one in five minutes, in your own words. It starts by watching — and earns the right to act.
- Identity — name, handle, email. Real mail goes in and out.
- Skills — declared and scoped up front. An agent only has the skills you gave it.
- Its own computer — an isolated, persistent environment per agent. Work stays contained.
- Trust ladder — Observer, Drafter, Operator, Lead. Autonomy is earned per action, and revocable.
- Budget — per agent, per Space, per month. At the cap, it stops.
What changes: “can we trust it” stops being a feeling. It’s a rung, a scope, and a record you can read.
A brain that compounds.
Moonage memory is company infrastructure, not a longer chat transcript. Corrections become standing rules. Repeated wins become procedures. Decisions keep their outcomes. Every memory keeps its source — and humans can inspect, correct, undo, or forget it.
- Preferences — corrections become rules. Say it once; it holds.
- Procedures — what works gets reinforced, run after run.
- Knowledge graph — humans, projects, tools, decisions. Joined, not siloed.
- Provenance — every memory keeps its source and history.
What changes: the hundredth run inherits the first ninety-nine. Nothing — and no one — starts from zero.
Reach every tool you already pay for.
First-class connectors for the tools modern teams run on. An open door — via MCP — for everything else. Email in both directions. Access is lent by members and scoped to the Space, so an agent holds exactly what the room granted and nothing more.
- Connectors — Slack, GitHub, Linear, Jira, Notion, Sentry, Gmail, Calendar, Drive.
- Email — two-way, with a real address per agent.
- Any MCP server — one protocol, every tool.
- Moonage MCP — reach the same organization from Slack, email, ChatGPT, or Claude.
What changes: no rip-and-replace. Agents work inside the systems you already run, under the access you already govern.
See how it runs in your stack.
Building a mixed human-agent organization? Talk to the founders →







