Where things live.

An agent has no screen of its own. Its work, its schedules, its documents and its settings all live in its team. Learn the shape once and it holds for every agent you install.

Your teams, in the sidebar

The sidebar is the whole map. Under Your teams, each team is a block of four rows, then the agents that belong to it:

• Mission Control. The team's board of work.
• Routines. What its agents do on their own.
• Files. What its agents keep.
• Team Settings. The team's agents, and the way into each one's settings page.

Every agent belongs to exactly one team, so there is no loose remainder at the bottom of the list. A new agent lands in the default team, which wears your workspace's name; drag it into another team to move it. Click an agent and you land on its team's Mission Control, already narrowed to that agent, and the narrowing follows you when you click Routines or Files next.

Team Settings appears for people who can configure something in that team. On your own computer that is always you.

Mission Control

A kanban board of missions in three columns: Running, Needs you, Done. Each card is one conversation with one agent. Click a card and the chat opens beside the board.

The board covers the whole team at once, which is the point: you watch four agents working without visiting four screens. The dropdown in the toolbar narrows it to one agent when you want that instead.

A finished mission sits in Needs you until you sign it off, then in Done until you archive it. Drag cards between those two columns, or use the one button on the card. Archived missions move to the Archived view behind the button in the toolbar, with its own search; "Back to missions" returns you. Reply to an archived mission and it comes back to the active board on its own.

The Mission Control entry at the top of the sidebar, above your teams, is the same board across every team at once.

Routines and Files

Routines is everything the team's agents do without being asked, in one list, each row naming the agent that owns it. Schedules like "every Monday at 8am, summarize last week's inbox" and event triggers share the list, because the difference between them is an engineering detail, not something you should have to think about. A routine you are still setting up in chat gets a row too, so a half-built one is never lost when you close its chat.

Files is what the agents have read or written: drafts, notes, structured data, exports. Folders nest, so merging several agents' folders would invent a filesystem nobody has. This section shows one agent's tree at a time and you pick the agent from a dropdown. Browse it as a card grid or as a list.

Agent settings

Everything configurable about one agent lives on one page. Open Team Settings and click the agent's row. The page has a rail on the left with two groups.

Context, what the agent knows:

• Job description. The agent's CLAUDE.md. This is what it reads at the start of every conversation to know who it is and how to behave. Edit it whenever the role changes.
• Memory. What the agent has learned while working, as cards you can read and remove.

Permissions, what it and its team may reach:

• People with access, either everyone on your team or a list you pick.
• Apps, which of your connected integrations this agent may use.
• AI models, which models it may run on.
• Skills, the skills installed on it.

On a single computer the page is Job description, Memory and Skills. The three access rows appear once you share a workspace with other people, since there is nobody to grant access to before that. If you administer a shared workspace, the same page is also reachable from Settings, under Permissions.

Integrations

One place, the Integrations entry in the sidebar. Connect Gmail, Calendar, Slack, Notion and around a thousand more.

A connection belongs to you, not to one agent, so you connect Gmail once and every agent you run can be given it. What is per agent is which of those connected apps it may reach, and that is the Apps row on its settings page.