The Agent Group node enables you to compose multiple published agent applications into a coordinated group. A decision-making model automatically plans the execution workflow, selects, and orchestrates the required sub-agents based on task requirements. The decision model intelligently determines which sub-agents are needed — it does not necessarily invoke all sub-agents in the group. This node is ideal for scenarios requiring multi-role collaboration or intelligent task decomposition.
Usage Example
The Agent Group node automatically plans the execution flow based on task requirements and orchestrates sub-agents to process tasks collaboratively.
Adding the Node
Method 1: Add via the + Button
-
Hover your mouse over any node on the canvas (e.g., the Start node); a
+button appears automatically on its right side. -
Click the
+button, then select Agent Group from the pop-up node list. The node is added to the canvas and automatically connected to the upstream node.
Method 2: Drag & Drop from the Node Library
- In the left-hand node library, locate and select Agent Group.
- Drag the node onto the canvas at your desired location, then drag a connection line from the edge of an upstream node to this node.
Core Configuration
Input
The user’s request or task description. The decision model analyzes intent, plans the execution workflow, and schedules appropriate sub-agents accordingly. Select an upstream node’s output variable using the dropdown menu in the format NodeName/VariableName, e.g., Start/query.
Model Configuration
Select the decision-making model responsible for analyzing user input, planning the task execution flow, selecting suitable sub-agents, and orchestrating their execution. Note: This model does not generate final answers — it only decides which sub-agents to invoke.
Model Selection
Choose a large language model (LLM) from the dropdown selector (e.g., Qwen series models). If your target model isn’t listed, click More Models, then search for and select it in the pop-up model selection dialog.
Model Parameters
Click the gear icon (⚙️) next to the model selector to open parameter settings. Enable the toggle first, then configure the following:
| Parameter | Default Value | Description |
|---|---|---|
| Max Output Length | 1024 | Maximum number of tokens the model may generate in its response. |
| top_p | 0.80 | Controls output diversity — higher values yield more varied results. |
| temperature | 0.70 | Controls output randomness — higher values yield more stochastic outputs. |
| enable_thinking | Enabled | Enables deep reasoning mode; outputs step-by-step reasoning traces. |
| thinking_budget | 4000 | Maximum number of tokens allowed for chain-of-thought reasoning output. |
| result_format | message | Format of the returned result. |
| enable_search | Disabled | Enables web search capability. |
Group Name
A name identifying the purpose of this agent group. Required field; maximum 20 characters.
Agents
Select and add already-published agents from My Applications to this group.
Each agent requires a description (max. 200 characters), explaining its capabilities and applicable use cases. This description guides the decision model during scheduling — its quality directly impacts scheduling accuracy. Clearly articulating functional boundaries helps the decision model select the correct sub-agent(s) for specific tasks.
Node Outputs
| Variable Name | Type | Description |
|---|---|---|
result | Object | Output result of the agent group. |
result.agResult | String | Final textual output after sub-agents complete the task. |
result.agProcess | String | Structured Markdown-formatted reasoning trace, including full planning and execution logs. |
| agProcess Structure Explanation: | agProcess records the entire reasoning process in structured Markdown, enabling downstream nodes to parse the reasoning steps — or supporting debugging and auditing of agent scheduling decisions. | Section |
| Content Description | User Goal | Original user request input. |
| Planning Steps | Sequence of sub-agent invocations planned by the decision model. | Execution Details |
| Execution Process & Output per Sub-Agent | Execution Result | Final aggregated result (identical to agResult). |
Frequently Asked Questions
Does the decision model always invoke all sub-agents?
No. The decision model intelligently selects only the sub-agents necessary for the given task. For example, if your group contains three sub-agents, but a particular task can be fully resolved by just one, the model will automatically skip the irrelevant ones.
What’s the difference between the Agent Group node and the Agent Creation node?
- Agent Group node: Orchestrates multiple already-published agent applications, with automatic planning and task delegation handled by a decision model.
- Agent Creation node: Dynamically creates a new, temporary agent within the current workflow. It is not published as a standalone application and exists only inside the canvas.