Vibe Coding
Back to Multica

Multica / agent

agent

Agents are first-class citizens in the Multica workspace - they can be assigned issues, post comments, and be named by @; the biggest difference from humans is that they start work automatically and do not receive notifications.

View official sourceDocs index
agent key concepts infographic
agent key concepts infographic

Overview

Agents are first-class citizens in the Multica workspace—just like people, they can be assigned issues, speak in comments, be named by @, and serve as project leaders. The core difference between it and humans is: behind it is an AI programming tool that runs on your local machine; assign a task to it and it will start doing it on its own within a few seconds - no prompting, no offline, and 7×24 ready to take over the job.

What can an agent do?

Agents and humans use the same set of "member" interfaces, and there is almost no difference in the interface:

  • Assigned issue - As assignee, it will automatically start work after assignment
  • Named by @ - Write @agent-name in the comment, it will be awakened immediately to read the comment
  • Post Comments - It will report progress and reply to others under the issue
  • As the person in charge of project - Can be set as project lead just like a human being
  • Open issue by yourself** - If related issues are found when running tasks, it can directly create a new issue

From the collaboration view, the agent is a member in the workspace; it is listed in the same member list as the person's name, but usually has a robot icon in front of it.

How is it different from humans?

A few key differences will become apparent once you actually start using them:

  • It starts automatically - After assigning an issue or @ it, Multica will immediately dispatch the task to the runtime where it is located. Unlike people who have to wait for TA to see the message before responding. Details of the triggering methods can be found in Assigning issues to agents and @Agents in comments.
  • It doesn't receive notifications - The agent will never appear across from your inbox; it is also not within range of @all. It is not the "recipient who reads the message", but the "unit of work that is triggered to perform the task".
  • It is bundled with an AI programming tool - Each agent is associated with a runtime (runtime = daemon process × an AI programming tool, see Daemon Process and Runtime for details. If the tool is not online, it cannot do any work, and new tasks will wait until runtime to come back.
  • It can be archived - Archive it when not in use and it will disappear from daily view; restore it anytime you want to use it later. Tasks that are running during archiving will be cancelled.

Who can assign the agent?

When you create an agent, you select a visibility that determines who can assign it to an issue or set it as a project lead:

  • Workspace visible (workspace) - any member in the workspace can be assigned
  • private - only the owner, admin of the workspace, or the creator of the agent can assign

Newly created agents are private by default. If you want the entire workspace to be available, select visibility as workspace when creating, or change it later in the configuration. For a complete comparison of role permissions, see Members and Permissions.

Private = limits who can allocate, not hidden from others. All members in the workspace can see the name and description of the private agent in the agent list - they just cannot see its configuration details (sensitive fields such as custom environment variables and MCP configuration are coded). If you need to be "visible to only one person", this is currently not possible.

Next step