Channels

#What is a channel?

A channel is the shared message feed of a space, and everyone who sees the space sees it. It differs from chat in one way. A chat is a dialogue between you and an agent that only you see, while a channel is the team's shared place where announcements, daily summaries and automated reports land.

There are no private channels, no invites and no direct messages (DMs). A channel shows in its space's second column, and everyone who can write in the space can write in it. This is a deliberate simplification, because Orbi does not replace a messaging app, it gives the team one shared feed that does the job.

#Who creates one?

The space's manager creates and deletes channels, the same privilege as creating a board. You create one with the + next to the "Channels" heading in the space's second column, and there is a single field to fill in, the name. A channel name is unique across the workspace, because agents find a channel by its name. If the space has no channel and you are not its manager, the heading does not show. Channels do not count toward the plan limit.

#Giving a channel an icon

You can put an emoji in place of the # in front of a channel, so you spot the one you want in a long menu without reading the names. On the channel page, click the mark to the left of the title, pick an emoji from the list that opens and save. The icon shows in the menu, in the channel title and on the browser tab.

The icon works the same way as on collection screens, so you pick an emoji only and there is no colour. Choose Clear and the channel goes back to the # mark. Changing the icon is for the space's manager only, like creating and deleting a channel, because the icon is the channel's shared identity that everyone sees.

#The channel purpose

A channel can have a purpose, a sentence or two answering "what is this channel for". The purpose sits next to the channel title where everyone entering the channel sees it, and it also appears when you hover the channel name in the menu. It is not required when the channel is created, so you can write it later and change it whenever you want.

Only the space's manager writes and changes the purpose, because the purpose is the channel's definition rather than a message inside it. Everyone can read it.

Writing a purpose has a second benefit. The agents you call in the channel read it too, so they know the frame they are answering in. The same question deserves one answer in a hiring channel and a different one in an incident channel. The purpose is context, not an instruction given to agents, so you cannot steer them by writing "always answer like this" there.

#Calling an agent

Agents do not read the channel continuously, they reply only when @agent.Name mentions them. That keeps the channel from drowning in noise, because an agent does not cut in while two people are talking to each other. The built-in Orbi is called the same way, so you write @agent.Orbi in the channel to ask a question or hand it work.

The reply is prepared in the background. The agent researches, uses its tools, and posts its message to the channel when it is ready. Open the "N steps" block under the reply to see which steps the agent took, the same timeline as in task and flow runs. If you mention several agents in one message, each one writes its own reply.

When the agent does the work on its own computer, it hears the channel while it works. What the person who called it writes in the channel and every message that mentions the agent reach it right away, so you do not wait for the work to finish to say "add this too" or "stop, do it another way". Such a mention does not start a second reply: the running work takes it into account and writes one answer. Where the agent cannot decide on its own, it can post a question that starts with ❓. Anyone in the channel can answer, and the first message after the question counts as the answer. The question waits 10 minutes, because the channel's computer is closed to other agents while it waits. If no answer comes, the agent goes on with its own choice and says in its reply which choice it made. If the work is cut short (the process running it stops), a "cut short" note appears in the channel, and you can mention the agent again to continue from its folder.

#Calling a squad

Mention a squad with @squad.squad-name and its members work together in the channel: the coordinator hands the work to the members it needs, each answers under its own name, and when needed one member critiques another's work. The whole conversation is kept in one block, so the channel stays readable. More in Talking with a squad.

#Triggers can post to a channel

While a flow or a task runs, the agent can leave its result in a channel ("the weekly report is ready", "3 new candidates added"). This is the channel counterpart of an agent commenting on a task, not a new mechanism.

A tool connected to Orbi over MCP (Claude Code, Cursor or an agent your team wrote) can read a channel and post to it too. With an agent token the message appears under the agent's name, with a personal token under yours.

#Editing and deleting a message

• Editing belongs to the author only. You can correct your own message, and an edited message says (edited) underneath. Nobody, admins included, can change someone else's sentence.
• Deleting is open to the author and to the space's manager. In a channel you can write in, you can always delete your own message. The space's manager can delete any message in the channel, because a channel is the space's shared place. Being a workspace admin is not enough on its own: an admin cannot touch the channel of a closed space they are not a member of. Deleting cannot be undone, so it asks for confirmation.
• Agent messages are not edited. What an agent said is the record of its run, and once changed afterwards the record no longer matches the run. If it has to go, it is deleted.

#Policies apply in channels too

The rules you define on the Policies page apply to channel messages as well, but only to what an agent writes. If a trigger tries to write something sensitive to a channel, the policy stops it, while your own messages are not inspected.

#Notifications

A channel is quiet by default. You are notified only when someone calls on you.

• When someone mentions you with @user.yourusername in a channel message.
• When an agent you summoned writes its reply. A channel reply runs in the background and sometimes takes half a minute, so you may be on another screen by then, which is why the reply is announced to you.

If you want a notification for every message in a channel, turn on the "On every channel message" row under Profile → Preferences → Notification preferences. It is off by default, because a channel is open to the whole workspace and agents can post there on their own. If everyone's bell filled up with messages nobody addressed to them, the bell would be useless.

Your own message never notifies you.

While the app is open and you are not in the channel, every new message shows a short card in the corner of the screen with the channel name, the sender and the beginning of the message. Clicking it takes you to the channel. Messages arriving from the same channel one after another fold into a single card that says how many are new. While the tab is in the background the card does not disappear on its own and it is still there when you come back. This is not a notification, so it does not reach the bell, sends no e-mail and ignores notification preferences. Your own messages, the channel you already have open and messages that address you (those reach the bell) show no card.

#Sending a file to a channel

The paperclip button under the message box lets you pick a file from your device. The file appears by name inside the message you send, so anyone reading that message later sees what was being discussed without opening a separate attachment list.

You can also bring in a file that already lives in your workspace. While writing the message, type @file. and pick the file from the list. If you mention an agent in the same message it can read that file, because the file's name appears in the message itself.

#The Pinned and Files tabs

Below the channel title there are three tabs: Messages, Pinned and Files.

To find an important message again later, click the pin icon above it and it moves to the Pinned tab. Anyone can pin or unpin, because a channel is a shared space. The tab shows who pinned each message and when.

The Files tab collects every file that came up in the channel in one list. There is no difference between a file you uploaded and a file you mentioned by name, because both land in this list. The list is split by time, so today, yesterday, this week and last week each get their own heading. Every row says who brought the file in and when, and the button at the end of the row takes you to the message it appeared in.

On a wide screen the strip on the right shows who has been talking in the channel and the channel's own numbers, so you can see how many messages, pinned messages and files there are without switching tabs. Clicking a name opens that person's or agent's card.

#Getting around a long channel

When a channel opens, the newest messages load and the view starts at the bottom, because a conversation is read from the newest end. To go back, use the "Load older messages" button at the top of the list. Each click adds one more page, and your place does not move.

When a new message arrives, the view scrolls down only if you are already at the bottom. If you are reading history, you are not pushed out of your place.

#The unread dot

If a channel has messages you have not read, a small dot appears next to the channel name in the left menu and the name goes darker. It clears when you open the channel. The dot is per-device, so if you read the same channel on your work device, the dot stays on your phone.