Skip to main content

Agent Team（Beta）

Last updated 10/03/2026

Agent Team (Beta) organizes multiple Agents into a collaborative team that works on a task together in different execution modes.

Version: 6.1. Intended for: users who need multiple Agents to collaborate.

Search keywords: Agent Team, Team, multi-agent, team collaboration, Beta.

The images reuse existing operation examples, so the old navigation names or locations in the screenshots may differ from the 6.1 interface. Locate the entry based on the text on this page and the current interface.

Agent Team (a team of AI agents) orchestrates multiple Agents to complete complex analysis tasks together. You can combine multiple specialized Agents into a team that breaks down tasks, processes them collaboratively, and summarizes the results in Leader–Worker mode.

Create and manage teams​

  • Quick creation from templates: The platform has several built-in team templates (such as Analytics and Engage, and Data Insights). You can pick a template to create a team in one click, or start from scratch with a custom configuration
  • Team configuration: Each team consists of several Agent steps. Each step is bound to one Agent, and you can set a Scenario Prompt and other options for it
  • Team List: On the team list page, you can view all teams (name, number of member Agents, and creation time). Hover over a team to preview its list of member Agents. Keyword search is supported

Execution modes​

  • Leader–Worker mode: A Leader Agent dynamically plans the task and dispatches other Worker Agents. After analyzing the input, the Leader decides on its own which Agents to dispatch and what to pass to them.

Task runs​

  • Start a run: After you enter a task description on the conversation or team page, the system automatically creates a TeamRun and pushes its execution progress in real time
  • Structured outputs: After each run completes, the system automatically generates structured outputs, including stage analysis reports, evidence chains, final reports, and run materials. They're displayed as cards grouped by Worker and stage
  • Quality validation: The system runs automatic quality checks on key outputs, and you can view the validation conclusion and detailed explanation for each output
  • Execution log: You can view the complete execution log of a TeamRun to troubleshoot abnormal steps

Scheduled automation tasks​

  • Create a scheduled task: Configure a scheduled task for any team, set its schedule (daily, weekly, monthly, and so on), and set the task message that's automatically sent to the team when the task is triggered
  • Status management: Scheduled tasks can be enabled or disabled. The management page shows the task name, last run time, and last run status
  • Manual trigger: You can manually trigger a scheduled task to run once right away. If the task already has a run in progress, the system rejects the duplicate trigger and returns a conflict message
  • Feishu notifications: You can configure a Feishu notification channel and target group. After a TeamRun completes, the system automatically pushes a notification with the run results to the specified Feishu group

Parent section: Agent. You can continue reading from the related pages above.

Was this page helpful?
On this page