All Products
Search

This Product

    DataWorks:New Data Agent

    Document Center

    DataWorks:New Data Agent

    Last Updated:Oct 09, 2026

    The new DataWorks Data Agent (based on Qwen Code) is now available on the public cloud. It provides a dedicated full-page interface, stronger reasoning, planning, and coding capabilities, and built-in DataWorks Skills for Data Integration, data development, data governance, and data security.

    Overview

    The new Data Agent is the next-generation AI agent from DataWorks, built on Qwen Code. Unlike the original Agent mode (a sidebar chat panel), the new Data Agent provides a dedicated full-page workspace for complex data development and governance tasks that require in-depth, sustained interaction.

    The new Data Agent provides the following core capabilities:

    • Stronger reasoning and planning: Built on Qwen Code, it understands complex data development and governance requirements and breaks tasks into multi-step execution plans.

    • Stronger coding capabilities: It generates, optimizes, and debugs high-quality code in SQL, Python, and other data development languages to help you produce executable code quickly.

    • Comprehensive built-in DataWorks Skills: Official Skills for Data Integration, data development, data governance, and data security are preset. You can use core platform capabilities without additional configuration. The built-in DataWorks ChatBI skill enables natural-language data querying and in-depth analysis within a session after it is enabled.

    • Conversational interaction: Chat mode provides an intuitive conversational interface. Describe your data development or governance requirement in natural language, and the agent interprets and carries out the task.

    Entry points

    You can open the new Data Agent in any of these ways:

    • Method 1: From the console navigation pane

      Log on to the DataWorks console and click Data Agent in the left navigation pane.

    • Method 2: From the All Products menu

      On the left side of Data Studio, click All Products. Under AI Agents, click Data Agent to open the new Data Agent.

    • Method 3: From the original Data Agent panel

      In the original Data Agent chat panel, click Try the New DataWorks Data Agent to switch to the new Data Agent.

    Quick start

    Step 1: Select or create an instance

    After opening the new Data Agent, select an instance with the status Running. Click the instance to open its Data Agent session page. If no instance is available, create a DataAgent instance as follows:

    1. On the instance list page, click Create Instance.

    2. In the Create DataAgent Instance dialog box, configure:

      Parameter

      Description

      Instance name

      A custom name that identifies the DataAgent instance.

      Workspace

      The DataWorks workspace associated with the instance.

      Resource group

      The Serverless resource group that runs the instance.

      CU

      The compute resource capacity required to run the instance. After you select a resource group, its available CU quota is displayed.

    3. Click Create Instance. After the instance starts and its status becomes Running, click it to begin using it.

    Step 2: Create a session

    After opening an instance, click New Task in the left sidebar to create a conversation. You can also click an existing session in the conversation history to continue it.

    Step 3: Switch the model and approval mode (optional)

    By default, Data Agent uses qwen3.7-max and Request Approval mode. You can switch the model or approval mode using the drop-down menus at the lower-left of the input box.

    Data Agent supports these models:

    Model

    Description

    qwen3.8-max

    The latest flagship Qwen model, with broad capabilities and a long context window. Suitable for complex data development and governance tasks.

    qwen3.7-max (default)

    The flagship Qwen model, strong at complex reasoning and code generation with balanced value for money.

    qwen3.7-plus

    A cost-effective Qwen model with fast responses, suitable for batch tasks and daily conversations.

    deepseek-v4-pro

    A DeepSeek reasoning model with strong deep reasoning, logical analysis, and coding capabilities.

    glm-5.2

    A Zhipu GLM model with strong understanding of Chinese semantics, suitable for general text processing.

    Data Agent provides these approval modes:

    Mode

    Behavior

    Use case

    Plan (plan)

    Analyze only; do not modify files or run commands.

    Understand the current state before deciding what to do.

    Request Approval (default)

    Ask for user confirmation before edits.

    Everyday use with a balance of efficiency and safety.

    Auto Edit (auto-edit)

    Automatically approve edits without step-by-step confirmation.

    Efficient development in a trusted environment.

    Smart Approval (auto)

    The agent assesses operation risk and requests user confirmation only for high-risk operations.

    Automated development with a balance of efficiency and safety.

    Full Access (yolo)

    Execute all tool calls and commands automatically without confirmation.

    Fully automated scenarios where you have sufficient confidence in the results.

    Step 4: Configure and use Skills (optional)

    Data Agent uses Skills to extend its AI capabilities. Use the built-in official DataWorks Skills or upload custom Skills to package business workflows and analytical methods for reuse.

    Upload a custom Skill

    1. In the Expand area of the session sidebar, click Skills to open the Skill management panel.

    2. Click Install Skill in the upper-right corner. In the dialog box, configure:

      Setting

      Description

      Name

      A unique identifier for the Skill. Use English letters and hyphens, for example, excel-analysis, so that you can refer to it in conversations.

      Description

      A description that helps users understand what the Skill does and when to use it.

      Skill body

      The main Skill content, written in Markdown. Define applicable scenarios, workflow steps, and output formats. Data Agent follows the instructions in the body when carrying out tasks.

      Zip package

      You can upload a resource package in .zip format, up to 50 MB per file. The root of the zip package may contain a SKILL.md file, which overrides the Skill body above when present. Other files are resource attachments and can be referenced by relative path from SKILL.md. Limits: SKILL.md ≤ 200 KB; entire zip package ≤ 50 MB.

    3. When configuration is complete, click OK. The uploaded Skill appears on the Custom tab of the Skill list.

    Use a Skill in a conversation

    Click the button below the conversation input box, imagethen select Skills from the context menu and choose the Skill to apply to this conversation. Data Agent follows its workflow and instructions.

    Note

    For details about managing Skills, see Data Agent Settings.

    Step 5: Enter an instruction and run it

    Describe your data development or governance requirement in natural language in the input box and press Enter. Data Agent plans and carries out the task using built-in DataWorks Skills.

    Example instructions:

    • "Create a data synchronization task from MySQL to MaxCompute."

    • "The department field for employee EMP001 in my_project.ods_emp_info_d is empty. Investigate the cause and recommend a fix."

    • "Show the lineage of the dws_user_active_di table."

    • "Generate a SQL script to analyze user activity."

    Interactive confirmations

    Data Agent asks for your confirmation at key points while running a task to keep operations under control. In these scenarios, it pauses and waits for confirmation:

    Confirmation type

    Trigger

    Description

    Plan confirmation

    After the agent breaks down a complex task.

    Shows the execution plan and runs it after you confirm the direction.

    Destructive operation confirmation

    Before deleting, overwriting, or clearing data.

    For example, DROP TABLE, TRUNCATE, or overwriting a partition.

    Publish or production confirmation

    Before submitting code or tasks to production.

    For example, publishing a scheduled task, deploying a node, or submitting SQL to a production project.

    Additional information

    When the agent lacks key information.

    For example, an ambiguous table name, uncertain field type, or unspecified time range.

    High-cost operation confirmation

    Before a full table scan or large JOIN.

    Shows an estimated cost so that you can decide whether to proceed.

    Cross-boundary operation confirmation

    Before operations across projects, tenants, or environments.

    Prevents unintended crossing of environment or permission boundaries.

    Permission change confirmation

    Before granting authorization, sharing, or changing an ACL.

    For example, granting table access or sharing a dataset.

    Long-running task interruption

    At key points of a task expected to take a long time.

    Reports progress and lets you continue or stop.

    Integration with ChatBI

    Data Agent provides built-in official DataWorks Skills. The DataWorks ChatBI Skill enables natural-language data queries and in-depth analysis in a session. After enabling it, you can perform data analysis in Data Agent conversations and receive reports with visual charts. For descriptions of built-in Skills, see Extensions.

    Enable the DataWorks ChatBI Skill

    1. In the session sidebar, click Extensions > Skills to open the Skill management panel.

    2. and switch to the DataWorks tab to view the official built-in Skills.

    3. Find the DataWorks ChatBI Skill card and turn on its switch.

    Note

    Built-in official Skills are enabled or disabled per instance. For information about uploading and using custom Skills, see Step 4: Configure and use Skills (optional) in Quick start.

    Automatic routing to the ChatBI Skill

    After the DataWorks ChatBI Skill is enabled, Data Agent automatically routes questions about data analysis or data queries to it. ChatBI analyzes the target dataset and returns the results without requiring you to switch tools.

    Explicitly specify the ChatBI Skill

    To ensure that your question uses ChatBI analysis, explicitly specify the Skill name. Enter /dataworks-chatbi-skill and then your data question. Data Agent calls the DataWorks ChatBI Skill for that question.

    Note

    Before using the ChatBI Skill for data analysis, create a dataset in ChatBI and complete its configuration. See Overview and Datasets.

    Billing

    For pricing details of the new Data Agent, see Data Agent fees.

    Related documents