Agent configuration

Agent configuration defines an agent’s name, purpose, and working rules, and determines which knowledge bases, skills, and tools it can use. After creating an agent, you can review and update these settings in the agent details.

Open agent configuration

1. Enter the target workspace.

2. Open the Agents page.

3. Open the Agents panel on the right.

4. Select the target agent to open its details.

Configuration components

Setting	Purpose
Name	Identifies the agent in the agent list and conversations
Description	Explains the agent’s purpose, intended users, and capability scope
System prompt	Defines the agent’s role, task rules, behavior limits, and output requirements
Knowledge bases	Provide business knowledge the agent can retrieve and reference
Skills	Provide methods, steps, and rules for specific tasks
Tools	Let the agent query external systems or perform actions

Basic information and system prompt

The name and description identify the agent and explain its purpose and capability scope, helping users choose the appropriate agent.

The system prompt defines the long-term rules the agent follows, including its role, task requirements, behavior limits, and output format. Changes to the name, description, or system prompt apply to subsequent conversations and tasks.

Capability bindings

Knowledge bases, skills, and tools are added to an agent through bindings. After binding a resource, the agent can use that capability when running tasks. Unbinding removes only the association between the agent and the resource; it does not delete the resource.

Resource availability also depends on the workspace, the current user’s permissions, and any authorization required by tools. For instructions, see Knowledge base binding, Skill configuration, and Tool configuration.