AI Features
1. KodeFlex Introduction
1.1 About KodeFlex
2. Installing KodeFlex
2.1 Windows One-Click Installer
2.2 Linux One-Click Installer
2.3 Installing ZAI Service
2.4 Installing from the Source Package (All Operating Systems)
3. Upgrading KodeFlex
4. Platform
4.1. Application Management
4.1.1 Create an Application
4.1.2 Maintenance Application
4.2. Module Design
4.2.1 About the Module
4.2.2 Main Table Design
4.2.3 Sub-table Design
4.2.4 Action Design
4.2.5 Label Design
4.2.6 More Settings
4.2.7 Release Deactivated Modules
4.3. AI Features
4.3.1 Intelligent Conversation
4.3.2 Agent
4.3.3 Skill
4.3.4 Model List
4.4. Application Instance Management
4.4.1 Maintenance Instance
4.4.2 Navigation Menu
4.4.3 Data Source
4.4.4 Validation Rules
4.4.5 Approval Process
4.5. User
4.5.1 Maintain Users
4.5.2 Maintenance Department
4.6. Permissions
4.6.1 Maintenance Permissions
4.7. System
4.7.1 Company
4.7.2 Backup
4.7.3 Authorization Information
4.7.4 Safety
4.7.5 Timing
4.7.6 Time zone
4.7.7 LDAP
5. Business Applications
5.1 Access Applications
5.2. Backend
5.2.1 Personnel Management
5.2.2 Access Management

Agent

2026-07-30 16:03:00
haoyonghong
179
Last edited by haoyonghong on 2026-08-05 16:02:28
Share links

Catalog

1. What is an Agent

Agents in the ZenTao Intelligent Development Platform act as AI assistants within your applications. They can help you with various tasks, such as automatically generating task descriptions or creating tasks based on requirements. 

You don't need to write any code. Simply configure the agent step-by-step in the interface to create a dedicated AI assistant. Once configured, you and your team can just click a button on the corresponding page while using the application, and the agent will handle the task automatically. 

For example, when creating a bug, you only need to enter a few keywords. The agent will automatically complete the bug description and reproduction steps based on this information, saving you a significant amount of manual typing time.

2. Who Can Design Agents

Designing and configuring agents requires administrator-assigned permissions. If you have been granted the necessary permissions, you can find the Agent feature in the application's AI module to create and configure your own agents.

3. How to Design an Agent

Navigate to the agent list page of an application and click the "Create Agent" button. The system will guide you through a five-step configuration process: Basic Information -> Set Input Fields -> Set Input Form -> Set Prompt -> Result Preview.

Step 1: Basic Information

This step tells the system who the agent is for, what it does, where it appears, and what it's called. 


  • Target Object (Required): Which functional module's data should this agent process? Select from the dropdown list (e.g., Requirements, Tasks, Bugs). 
  • Action Purpose (Required): What type of operation should the agent perform? Options such as "Create" or "Edit" will automatically update based on the Target Object you selected. 
  • Display Location (Required): Where should the agent appear? Choose either the Detail View page or the Form Entry page. 
  • Name (Required): Give your agent a name. It defaults to the Action Purpose name, but you can customize it as needed. 
  • Model (Required): Select the AI model powering the agent. The default model is usually sufficient. 
  • Description (Optional): Provide additional details about the agent's functionality to help colleagues understand its purpose.

Step 2: Set Input Fields

This step determines what background information to provide to the AI. Based on the Target Object selected in Step 1, the system will list all available fields for the AI to reference. Simply check the boxes for the fields you want to include. 

Note: If you selected "Form" as the Display Location in Step 1, this step will be skipped automatically. On form pages, the system directly reads the current form fields as context. 

For example, if you're creating an agent to write bug descriptions and selected "Bug" as the Target Object, you can check fields like Bug Title, Severity, and Module. The AI will then reference this information when generating the description.

Step 3: Set Input Form

Sometimes, the object's native fields aren't enough, and you need users to provide additional context when invoking the agent. This step is for defining those supplementary input fields. 

For example, if you're building an agent that creates tasks from requirements, you might want users to specify the Task Type (Development, Design) or Priority in addition to the requirement's native fields. These extra options can be configured here. 

You can add various form field types, such as text boxes, dropdown menus, and date pickers. When users invoke the agent, they will see these input fields. The agent will then combine this information to generate more accurate results.

Step 4: Set Prompt

This is the agent's "brain"—where you define how the AI thinks and responds. 


  • Role Definition (Optional): Tell the AI what role to play, such as "You are a senior project manager." Setting a role makes the AI's responses more professional and contextually appropriate. 
  • Prompt Content (Required): The core instructions telling the AI exactly what to do and how to format its output. This is the most critical part of the agent configuration.
  • Attached Skills (Optional): Attach published public skills to equip the agent with additional capabilities.


Step 5: Result Preview & Publish

This is the final step. The system summarizes the configurations from the previous four steps, allowing you to review everything before publishing. 


  • Debug: Test the agent in debug mode without publishing it first. Once you're satisfied with the performance, return here to publish. 
  • Publish: If everything looks correct, publish directly. Once published, team members will be able to see and use the agent on the corresponding pages. 


We recommend running a few debug tests to ensure the output quality meets your expectations before publishing. You can always return to modify the agent after it has been published.

4. Using Agents in Different Scenarios

When configuring an agent, you can choose between two display locations: Form or Detail.

Using Agents on Form Pages:

After setting up an agent, an Agent button will appear in the top-right corner of the form when creating or editing an item (e.g., creating a new task). 

Here's how to use it: 

Trigger the Agent: Click the Agent button in the top-right corner of the form. If multiple agents are available, a selection list will appear; if only one is available, its name will be displayed directly. 

Agent Execution: Once clicked, the system automatically reads the relevant fields of the current form object, any content you've already entered, the supplementary input fields configured in the form, and any attached skills. 

Preview Results: The AI-generated results will be displayed in a panel for you to preview. 

Populate Form: After confirming the results, click the "Populate" button to automatically fill the corresponding form fields. 

The agent is designed to assist with form filling and will not interfere with the normal form submission process. You can still fill out forms entirely manually, or let the agent fill in part of the form and then make manual adjustments.

Using Agents on Detail Pages:

When viewing an object's detail page (e.g., Task Details), the corresponding agent entry point will be displayed. Simply click to invoke it. 

Here's how to use it: 

Trigger the Agent: Navigate to the detail page of an object that supports agents (e.g., Task Details), locate the agent entry point on the page, and click the corresponding agent name. 

Agent Execution: Once clicked, the system automatically reads the field information of the current object as configured in the agent, the supplementary input fields configured in the form, and any attached skills.

Preview Results: The AI-generated results will be displayed in a panel for you to preview. 

Populate Form: After confirming the results, click the "Populate" button to automatically fill the corresponding form fields.

Write a Comment
Comment will be posted after it is reviewed.