The Data Agent settings center is the core module for configuring and optimizing the intelligent assistant. It currently integrates core functions such as MCP Servers, Rules, General, and Usage. By connecting to external tool services and enabling flexible Rule customization, Data Agent improves development efficiency, ensures code quality, and provides a highly personalized and intelligent development experience.
Access the settings center
The Data Agent settings center allows you to configure Rules, MCP Servers, Usage, and General settings. To access the settings center, follow these steps:
In the upper-right corner of the DataWorks UI, click the Data Agent icon to open the Data Agent panel.
On the Data Agent panel, click the
Settings icon to go to the settings center.
General
In the Data Agent settings center, click Settings to open the Settings tab. On this page, you can configure the following personal setting:
The default storage path for generating code files: Specifies the storage path for code generated by the Agent. The default option is Project Directory. You can switch it to Personal Directory as needed.
Seat management
This feature is available only in Data Agent Team Edition. For more information about Data Agent editions, see Data Agent billing.
Seat overview
In the Data Agent settings center, click Seat Management.
Metric | Description |
Token quota | The token quota provided for each seat in the Team Edition. |
Number of seats | The total number of seats you have purchased. You can click Upgrade in the upper-right corner to adjust the number of seats. |
Allocated seats | The number of seats that have been allocated to users. |
Remaining allocatable seats | The number of available seats that have not been allocated. This is the difference between the total number of seats and the number of allocated seats. |
Allocate seats
If there are remaining allocatable seats, a tenant administrator can allocate seats to specified Alibaba Cloud accounts. To do so, perform the following steps:
In the seat list area on the seat management tab, click Allocate Seat.
In the dialog box that appears, select the Alibaba Cloud account to which you want to allocate a seat.
Click Confirm to complete the allocation. The user will then appear in the seat list and can begin using Data Agent features.
Deallocate seats
A tenant administrator can deallocate seats to reclaim them. After a seat is deallocated, the corresponding user loses access to Data Agent, and the reclaimed seat can be reallocated to another user.
To deallocate a single seat: In the seat list, find the target seat and click Deallocate in the Actions column.
To deallocate multiple seats in a batch: In the seat list, select the checkboxes of the seats that you want to deallocate, and then click the Batch Deallocate button below the list.
Seat list
The seat list provides the following details for each allocated seat:
Field | Description |
Seat number | The unique identifier of the seat, which is automatically generated by the system. |
Owner | The user account to which the seat is allocated. |
Token usage | The number of tokens consumed by the seat. Administrators can use this metric to monitor resource usage for each seat. |
Allocation time | The time when the seat was allocated to the current user. |
Actions | For an allocated seat, you can click Deallocate to reclaim it. |
Usage
The Usage feature is a data insight tool for administrators and developers. It helps you quantitatively evaluate AI feature usage, track trends, and manage token consumption. In the Data Agent settings center, click Usage to open the Usage tab.
The Usage feature currently supports the following regions: China (Chengdu), China (Shenzhen), China (Beijing), China (Shanghai), and China (Hangzhou). Support for other regions will be available soon.
Core metrics
The usage report allows you to view core Data Agent usage data from different perspectives. You can use the filters at the top of the page to customize the analysis scope:
Statistical scope: Switch between different perspectives to view usage data.
Perspective
Description
Tenant perspective
View aggregated data for all workspaces under the current tenant.
Workspace perspective
View usage data for the workspaces that you have permission to access.
Personal perspective
View your personal usage data. If you are a tenant administrator, you can also view the usage of other members in the tenant. Non-administrator users can only view their own usage data.
Statistical period: Customize the start and end dates to view usage data for a specific time period.
Feature point: Filter data by feature module (multiple selections supported) to view the usage of features such as Agent, code programming assistant, quick AI actions, and ChatBI smart analysis.
After you select the statistical scope, statistical period, and feature points, the page displays the following three core metrics. Each metric includes its week-over-week change trend.
Number of requests: The total number of Data Agent requests that meet the specified conditions within the current statistical period.
Number of users: The number of unique users who used Data Agent and meet the specified conditions within the current statistical period.
Tokens consumed: The total number of tokens, including both input and output, consumed by Data Agent requests that meet the specified conditions within the current statistical period.
Below the core metrics, the Metric Details section displays a line chart that shows the trend of core metrics over the selected period. You can enable Group by feature point to compare the usage proportions of different feature modules. The Activity Index section uses a calendar heatmap to show usage activity throughout the year. Darker colors indicate higher usage on a given day. This helps you easily identify peak and off-peak periods.
Token consumption details
The token consumption details table records the resource consumption for each AI call. This helps you perform cost analysis, troubleshoot anomalies, and audit usage. The table contains the following fields:
Field | Description |
Time | The time when the request was made. |
User | The identifier of the user who initiated the request. |
request ID | The unique identifier for each request. If you detect abnormal token consumption, you can use the request ID to locate and investigate the specific call. |
Feature point | The feature module used by the request, such as code programming assistant or Agent. |
Intent name | The specific intent type identified for the request, such as single-line smart completion or data development Agent. |
Input tokens | The number of input tokens sent to the model in the request. |
Output tokens | The number of output tokens returned by the model. |
Total tokens | The sum of input and output tokens consumed by the request. |
Skills
A Skill is a custom instruction module in Data Agent that extends its AI capabilities. By creating a Skill, you can encapsulate specific business processes, data processing logic, or analysis methods into a reusable component. This lets Data Agent perform specialized tasks by following a predefined workflow, such as Excel pivot table analysis or SQL quality reviews.
Create a skill
In the Data Agent settings center, click the Skill tab. Click Create Skill and configure the following parameters:
Parameter | Description |
Name | The unique name for the Skill. We recommend that you use English and hyphens, such as |
Description | A description of the Skill's function. This helps users quickly understand its purpose and applicable scenarios. |
Skill body | The core content of the Skill, written in Markdown format. You can define applicable scenarios, workflow steps, and output formats. Data Agent executes tasks according to the instructions in the body. |
Upload file | You can upload a resource file package in |
Scope | Select the visibility scope for the Skill:
|
Use a skill in a conversation
After a Skill is created, you can use it in conversations with Data Agent. Below the conversation input box, click the 
icon. In the context menu that appears, select Skill, and then choose the specific Skill that you want to apply to the current conversation. Data Agent then executes the task according to the workflow and instructions defined in that Skill.
MCP Servers
An MCP (Model Context Protocol) Server is a collection of backend tool services that the Data Agent relies on to perform tasks. It provides tools, data sources, and APIs for operations such as queries, analysis, and code generation. In the settings center, you can view the official built-in Alibaba Cloud-DataWorks-MCP-Server and its related tools.
Access MCP Servers
In the Data Agent settings center, click MCP Servers to go to the MCP Servers tab.
Use an MCP server
MCP Server-related tools can be used in the Agent. For more information, see Data Agent.
Rules
A Rule is the core mechanism for providing persistent context, standards, and preferences to DataWorks Data Agent. This ensures that the code and responses it generates precisely follow your specific requirements.
Rule types and permissions
DataWorks provides two types of Rules: Personal and Enterprise.
Type | Definition and purpose | Permission control | Scope |
Personal Rules | Created and maintained by an individual developer to encapsulate personal coding habits, frequently used code snippets, and project descriptions. | Visible to and effective only for its creator. Other users cannot view or use it. | Applies at the personal level. Can be invoked in any workspace that the user can access. |
Enterprise-class Rules | Defined by an administrator to establish and enforce common development standards, such as data warehouse layering conventions, code style guides, and usage instructions for core tables. | Can be created, edited, and managed by workspace administrators and users with higher-level permissions. | Can be applied globally or to specific workspaces. |
The Enterprise-class Rules feature is available only in Data Agent Team Edition. For more information about Data Agent editions, see Data Agent billing.
Create and manage rules
In the Data Agent settings center, switch to the Rules tab. On the Rule management page, you can perform the following actions:
Switch between the Personal Rules and Enterprise-class Rules tabs to manage each type of Rule separately.
View information about existing Rules, such as their name, activation mechanism, and scope.
View, edit, or delete existing Rules.
Click New Rule to create a new Rule.
When you create or edit a Rule, configure the following core attributes:
Attribute | Description |
Rule Name | Set an easily recognizable name and a detailed description for the Rule. |
Rule Content | The core of the Rule, which is the specific contextual prompt information that you want to provide to the AI. You can specify the standards and conventions that Data Agent must follow when generating code. Both manual input and document uploads are supported.
|
Entry into force mechanism | Defines the policy for how the Rule is automatically introduced into a conversation. This aligns with the design philosophy of Cursor.
|
Scope of entry into force | For an Enterprise Rule, you can set a scope to control where the Rule is visible and usable.
|
The following content is an example of a Rule:
Rule name: Table and node naming conventions
Rule content:
# DataWorks data warehouse (ODS/DWD/DWS/ADS) table and node naming conventions As a senior data warehouse architect, you must strictly follow these naming conventions when you create any table or node in DataWorks. These conventions are essential for ensuring that your data assets are clear, maintainable, and consistent. ### 1. Table Naming Convention All table names must follow a unified, structured paradigm. #### 1.1 Core naming formula LayerPrefix_[CustomDescription]_[UpdateStrategySuffix] #### 1.2 Layer Prefix - [Mandatory] - DIM (Dimension layer): Must start with dim_. - DWD (Detail layer): Must start with dwd_. - DWS (Summary layer): Must start with dws_. - ADS (Application layer): Must start with ads_. #### 1.3 Update Strategy Suffix - [Mandatory] - DIM (Dimension layer): Use the _df suffix, which represents a Daily Full Snapshot. - DWD (Detail layer): Use the _di suffix, which represents Daily Incremental data. Use the _df suffix, which represents a Daily Full Snapshot. - DWS / ADS (Summary/Application layer): Use suffixes such as _1d, _7d, and _nd to represent the data aggregation period (for example, last 1 day, last 7 days, or last N days). #### 1.4 Delimiter All words in a table name must be in lowercase and separated by a single underscore (_). #### 1.5 Naming examples - DIM table example: dim_user_info_df (User information dimension table, daily full) - DWD table example (incremental): dwd_trade_order_detail_di (Trade order details, daily incremental) - DWD table example (full): dwd_product_base_info_df (Product basic information, daily full) - DWS table example: dws_user_active_uv_7d (User active UV summary for the last 7 days) - ADS table example: ads_screen_kpi_overview_1d (Dashboard core KPI overview, daily) ### 2. Node Naming Convention The node name should clearly reflect the core table it produces and its processing logic. - Naming principle: The node name should be highly consistent with the name of the main table it produces, in the format [Layer]_[BusinessLogic]. - Example: - An ODPS SQL node that produces the dwd_trade_order_detail_di table should be named dwd_trade_order_detail_di. - If a node handles complex logic, it can be named dws_build_user_active_uv_7d.
Use rules in conversations
Configured Rules take effect during your daily interactions with Data Agent.
For Rules that are set to Always applied., no extra action is required. When Data Agent generates code or responses, it uses these Rules as background knowledge and constraints by default.
For Rules that are set to Manual application, you can manually enable them during a conversation:
Click the
button below the input box.In the context menu that appears, select Rule, and then select the specific Rule that you want to apply to the current conversation.
FAQ
Difference between skills and rules
Both Skills and Rules are mechanisms for extending the capabilities of Data Agent, but they have different roles and purposes:
Dimension | Skill | Rule |
Role | A task-oriented, reusable workflow that encapsulates specific business processes and execution steps. | A constraint-oriented, persistent context that provides the AI with standards, preferences, and background knowledge. |
Purpose | Defines "what to do" and "how to do it" for specific tasks, such as Excel pivot table analysis or SQL quality review. | Defines "what standards to follow," such as data warehouse naming conventions, code style guides, or usage instructions for core tables. |
Activation method | On-demand invocation. You trigger a Skill by manually selecting it in a conversation. | Supports both automatic activation (always-on) and manual activation (on-demand). |
Attachment support | Supports the upload of a .zip resource package, which can include data files, examples, and other resources. | Supports manual input or the upload of document files, such as .doc, .pdf, or .md files, as Rule content. |
Recommendations:
Use a Skill when you need Data Agent to complete a specific task with a fixed set of steps, such as data analysis or report generation.
Use a Rule when you need Data Agent to consistently follow certain standards, such as naming conventions or code styles, in all conversations.
Skills and Rules can be used together. For example, you can use a Rule to define your team's coding standards and then use a Skill to encapsulate a specific code review process. This allows Data Agent to automatically follow the team's standards when it performs the review task.
Difference between skills and MCP servers
Both Skills and MCP Servers extend the capabilities of Data Agent, but they operate at different levels:
Dimension | Skill | MCP Server |
Role | Instruction layer. Defines a workflow and execution logic by using natural language to tell the AI "how to complete a task." | Tool layer. Provides the AI with callable external tools and data sources, giving it the "ability to perform actions." |
Capability scope | Orchestrates existing capabilities by combining multiple steps into a complete business process. | Extends new capabilities by integrating with external system APIs and data sources, such as by querying data or executing actions. |
Typical scenarios | Excel pivot table analysis, SQL quality review, and code review workflows. | Querying DataWorks metadata, calling external APIs, and accessing data sources. |
Recommendations:
Use a Skill when you need to orchestrate existing capabilities to complete a specific business process. You can create one quickly without programming.
Use an MCP Server when you need to enable Data Agent to access external systems by integrating with the corresponding tool services.
Skills and MCP Servers can work together. An MCP Server provides the underlying tool capabilities, and a Skill orchestrates those tools into a workflow. For example, an MCP Server can provide a tool for querying DataWorks metadata, and a Skill can use that tool to define a complete data lineage analysis process.