Chat with agents
A one-to-one conversation with an agent that only you can see. The agent starts with no issue context and doesn't create issues on its own; when you explicitly ask, it acts with your workspace access.
Chat fits work that hasn't become a clear issue yet: sorting out ideas, checking on the workspace, discussing an approach, asking a question that doesn't belong to any issue, or having an agent finish a small task first.
Chat is a one-to-one conversation between you and an agent, off the issue board. The whole conversation is fully private: nobody else in the workspace, admins included, can see it.
A chat is not bound to any issue. The agent isn't watching your board, doesn't automatically receive an issue's description and comments, and won't turn the conversation into issues on its own. It can still use the Multica CLI to query workspaces, projects, issues, and skills within its permissions, and when you explicitly ask, it acts with your workspace access, for example to create an issue.
A chat can also carry background. Click + at the bottom left of the input and choose Project context to attach a project to the conversation; when the agent runs, it receives that project's description, repositories, and other resources.
Start a conversation
Open Chat in the sidebar.
Click new and pick an agent you have permission to run.
Type a message; you can also attach images or files.
After you send, Multica creates a run for the message, and the reply comes back to the current conversation.
If the runtime is offline, the message waits until it comes back online. Once the agent is archived, or you lose permission to run it, you can't send new messages.
Before the first message, Chat shows up to three suggestions configured for the selected agent. Selecting one only fills the composer with its full prompt; nothing is sent, so you can edit it first. Agents without configured suggestions show general-purpose examples.
After the agent replies, it sometimes offers Follow-up questions below the reply. Clicking one sends it right away.
Handles office and operations chores: checks progress, plans schedules, files issues.
Chat from any page
You don't have to leave the page you're on to chat. Workspace pages can open a chat in a floating window. The window and the Chat page share the same conversations, so you can pick up a chat from either one.
In a wide window, quick chat sits at the right end of the status bar along the bottom of the app:
- Ask Multica starts a new chat.
- A chat gets a chip in the bar while the agent is working on it, while it has a reply you haven't read, and once you've opened it in the floating window. A spinner on the chip means the agent is replying, and a colored dot means an unread reply; hover the chip to see the chat's title and state. Click a chip to open that chat, and click it again to close the window. Chips that don't fit go into a +N menu.
- To close a chip, hover it and click ×. It comes back when the chat gets a new message or when you open the chat again.
- Chat history, the last button in the bar, lists your recent chats with the same running and unread marks. Pick one to open it, or click All chats to go to the Chat page.
- The window opens above the button or chip you clicked.
A narrow window has no status bar. Instead, a round chat button in the bottom-right corner opens the window. The button pulses gently while an agent is replying, and its tooltip says when a chat is running or has unread replies.
- Press
⌘J(Ctrl+Jon Windows and Linux) to open or close the floating window; opening it puts the cursor in the message input. The key works while you're typing in that input, too. You can rebind it under Settings → Shortcuts; see Search, shortcuts, and preferences. - Press
⇧⌘J(Ctrl+Shift+Jon Windows and Linux) to start a new chat instead, the same as Ask Multica;⌘Jreopens the chat you had open last. Press it again on the new chat to close the window. - Drag the window's edges to resize it, or expand or minimize it.
- On the Chat page itself, quick chat doesn't appear in the status bar and the round button doesn't show. Neither
⌘Jnor⇧⌘Jdoes anything there: the page already shows the same conversation.
If you don't want the floating window, turn off Floating chat window under Settings → Preferences → Comments & chat. Quick chat then leaves the status bar, the round button no longer appears, and Chat is reachable only from the sidebar. The switch applies to the current device only.
Chat vs. issues
| Need | Better fit |
|---|---|
| Quick questions, exploring an approach, or private drafts | Chat |
| A clear assignee, status, priority, and deliverable | Issue |
| Letting teammates see the background and continue the discussion | Issue |
| A quick look at workspace information with no work record needed | Chat |
Direct chats created in Multica can be read only by their creator, not by other members or admins. Put conclusions the team needs to see into an issue, a project description, or a skill.
Multi-turn context
A conversation tries to continue the original session in the AI coding tool, so later messages don't need to repeat earlier content. Multica stores the session ID and routes subsequent runs to a runtime that can access that session.
If the local session no longer exists, the context can't be continued, or the previous error shouldn't be reused, the runtime starts from a fresh session. The conversation's message history is kept either way.
Watch and stop a run
While the agent runs, the UI shows the current phase and tool activity. Expand View details to inspect the run in full.
To interrupt, click the stop button next to the input (shown while the input is empty). Content already returned to the conversation is kept.
If the agent hadn't produced anything yet, the message you just sent is taken back out of the conversation and returned to the input, attachments included, so you can edit and resend it. Stopping sometimes takes a moment to finish; refreshing the page or closing the window in the meantime doesn't lose the message, and it comes back to the input the next time you open that conversation. If the input already holds another draft, the message waits until you clear it.
Send messages while a reply is in progress
You can keep typing while the agent is replying. Send adds a plain text message without a skill link to the current reply without stopping it, when the runtime supports steering. The agent reads it at its next safe boundary and continues the original task. Upgrade your daemon to enable this.
On web and desktop, Add to current run above your draft means your text will join the running reply. Start after this run means the message will be queued, for example when it has files or a skill link, or the current run doesn't support steering.
Messages with attachments or skill links, and messages sent before a reply starts or to a runtime without steering support, enter the queue. If guidance is not delivered before the reply ends, it automatically becomes a queued message; you do not need to resend it. If you stop while delivery is in progress, Chat waits for its result before queuing undelivered guidance. If the reply is retried automatically, guidance stays with the original request so both are processed together. When resuming the same session, guidance already delivered there is not sent again.
Queued messages appear above the input and run in order. You can remove them, use ⋯ → Edit queued message to restore them to the input, or ⋯ → Clear all to withdraw them. Send now (stops current reply) moves a queued message to the front and stops the current reply. It requires an active reply and permission to run the agent.
Steering applies to direct Chat sessions on updated web, desktop, and mobile clients. On mobile, queued messages appear at the end of the conversation; remove or edit them on web or desktop. Older clients and Slack and Lark channel conversations keep their existing queue behavior.
Follow up later
Ask the agent to report back when something happens, for example, "Tell me when this issue is ready for review" or "Remind me in two hours." It can create a wakeup that starts another run in this conversation when an issue reaches a status, sub-issues finish, a GitHub PR merges or closes, its checks finish, a run ends, or a member comments on an issue. It can also return at a set time or check periodically.
On web and desktop, a card under the reply records the follow-up. The strip above the input shows what the agent is waiting for and its deadline. Open the strip or the clock button in the conversation header to see Wakeups, including those that ended in the past 7 days. Use Cancel wakeup to stop waiting; if a report run fails or is cancelled, use Retry to request it again.
Conditions and periodic checks wait 24 hours by default, at most 7 days. If the deadline arrives first, the agent gets one final turn to report where things stand. One-time reminders can be scheduled up to 30 days ahead. A conversation can have at most 10 active wakeups. Archiving the conversation cancels its active wakeups; losing access to the agent or watched work can also end them.
For ongoing schedules such as a daily report, use an autopilot. Agents manage follow-ups with chat wakeup.
Organize the conversation list
Conversations can be renamed, pinned, archived, and deleted:
- Pin keeps frequently used conversations at the top of the list;
- Archive moves a conversation to Archived and makes it read-only;
- Unarchive lets you send messages again;
- Delete permanently removes the conversation and its messages.
If a conversation is bound to an external channel (such as Slack), archiving breaks the binding; unarchiving does not restore it — bind again.
The number of unread replies across your conversations shows as a badge on Chat in the app rail. In the desktop app, right-click Chat for Mark all as read.
Deleting a conversation cannot be undone and stops any runs in it that haven't finished. Archive instead when you just don't need it for now.
Next steps
- Assign issues to agents — turn a discussion into trackable work.
- Inbox and subscriptions — follow team activity that needs you.
- Search, shortcuts, and preferences — open chat, search, and move between pages from the keyboard.
- Providers matrix — session and model capabilities per AI coding tool.