All Products
Search

This Product

    AI-Native Database Service:Analytics Agent

    Document Center

    AI-Native Database Service:Analytics Agent

    Last Updated:Sep 04, 2026

    Analytics Agent is an intelligent data analysis agent provided by AI-Native Database Service for enterprise users. It interprets natural language requirements, automatically understands your data, proposes analysis approaches, expands analytical perspectives, and delivers analysis results by invoking the appropriate tools.

    Scenarios

    • Business users: Define data requirements and quickly generate reports such as sales trends and user behavior analysis.

    • Data analysts: Perform exploratory analysis, delegate time-consuming tasks such as automated feature engineering and data cleaning, and reuse Python code.

    • Managers: Obtain in-depth insights and analysis on key business metrics to support data-driven decision-making.

    Features

    • Automated data insight exploration: Describe your requirements in natural language, and the agent autonomously understands the requirements, comprehends the data, generates insights, and produces reports.

    • Shortened data-to-value path: Focus on the relationship between data results and business decisions without worrying about complex data processing workflows.

    • Seamless enterprise data connectivity: Connect to ApsaraDB services and data sources managed through DMS for immediate data analysis.

    Extension and integration

    • Core component: Data Agent Skill

    • Quick integration: Pull and install the skill directly from the Git repository, or run the following command for quick deployment:

      npx skills add https://github.com/aliyun/data-agent-skill

    • Overview: This skill package can be downloaded and integrated into your agent service. It establishes a secure bridge between agents and enterprise data sources through standardized connection protocols, enhancing data access capabilities and analytical depth to accelerate the development of custom data-driven applications.

    Supported regions

    China: China (Beijing), China (Zhangjiakou), China (Ulanqab), China (Hangzhou), China (Shanghai), China (Shenzhen), China (Chengdu), and China (Hong Kong).

    Asia Pacific: Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Philippines (Manila), and Japan (Tokyo).

    Others: US (Silicon Valley) and US (Virginia).

    Account preparation

    Step 1: Create an account

    Create an Alibaba Cloud account for Data Agent. Create the required RAM users based on your business needs.

    Step 2: Grant permissions

    Grant the account permissions to use Data Agent. Policy name: AliyunDMSDataAgentFullAccess.

    Note

    This policy grants only the permissions to use Data Agent features. It does not include data access permissions. Data access permissions must be requested and approved separately during use.

    Authorization methods:

    • Method 1: Grant permissions to a single account

      1. Log on to the RAM user list by using an Alibaba Cloud account that has RAM administrator permissions.

      2. Find the RAM user that you want to authorize and click Actions > Grant Permission.

      3. In the dialog box that appears, search for and select the AliyunDMSDataAgentFullAccess policy, and then click Grant Permission.

    • Method 2: Grant permissions to a user group

      1. Log on to the RAM user group list by using an Alibaba Cloud account that has RAM administrator permissions.

      2. Select the target user group. On the user group details page, click the Permissions tab, and then click Grant Permission.

      3. Search for and select the AliyunDMSDataAgentFullAccess policy, and then click Grant Permission.

      4. Add the users that you want to authorize to this user group.

    Procedure

    1. Log on to the Agent console.

    2. Select an edition based on your business requirements. For information about the differences between editions, see Data Agent editions.

      • Free Edition:

        On the Free version card, click Free Trial.

      • Personal Edition:

        • On the Personal card, click Upgrade to Personal Edition.

        • Select the region, number of seats, and subscription duration based on your requirements.

          Note

          In the Subscription Duration field, click the monthly duration to select a period.

        • Click Buy Now and complete the purchase.

      • Enterprise Edition:

        • On the Enterprise card, click Upgrade to Enterprise Edition.

        • Select the region, number of large language model (LLM) resource packages, and subscription duration based on your requirements.

          Note

          In the Subscription Duration field, click the monthly duration to select a period from 1 to 11 months.

        • Click Buy Now and complete the purchase.

    3. Upload data samples by uploading from your local machine or selecting existing data.

    4. Wait for the agent to generate an execution plan, and then check whether the plan meets your requirements.

      • If the plan meets your requirements, click Start Task.

      • If the plan does not meet your requirements, click Modify Task to refine the requirement description until the analysis plan is satisfactory.

    5. Wait for the plan to be executed. During execution, you can view the code that the agent writes for each analysis step and the execution results.

    6. Have the agent generate a visual web report based on the data to present richer data insights.

    Data center

    In the left-side navigation pane, click Data Center, and then click Add Data. Select a data ingestion method based on the data source type.

    • Local Upload:

      Drag the files that you want to analyze to the upload area, or click the upload area to select files. After the upload is complete, click Confirm.

    • Cloud databases: The Free Edition and Personal Edition support RDS Database, PolarDB database, and AnalyticDB Database. The Enterprise Edition supports over 40 types of data sources. For the complete list, see Databases supported by DMS.

      1. Select the Region and Instance.

      2. Enter the Database Username and Database Password.

      3. Click Test link to continue.

      4. Select the Database and Tables.

      5. Click Confirm.

    • DMS-Managed Instances:

      You can use MaxCompute and Hologres instances that are already registered in DMS as data sources for analysis. For information about instance registration, see Register an instance.

    FAQ about adding data

    Databases and tables are not displayed after adding a MaxCompute instance

    MaxCompute instances purchased after 2024 require tenant-level Information_schema processing, and the architecture must be upgraded to the schema level. Perform the following steps:

    1. On the Project Management page in the MaxCompute console, perform the Upgrade to Support Schema Level operation.

    2. Grant the tenant-level admin permission to the user identity that corresponds to the AccessKey pair used when registering the instance in DMS.

    3. In DMS Instance Management, click Edit next to the target instance, and then click Save to trigger the instance to recognize the schema-level change.

    Connection error when adding a Hologres instance

    DMS requires a VPC endpoint to access a Hologres instance. Make sure that the Hologres instance has a VPC endpoint configured.

    For information about VPC endpoint configuration, see .

    Cannot find a database or table in the drop-down list

    Click the refresh button. If the number of tables is not large, the table typically becomes searchable within 30 seconds.

    Permission denied error when refreshing databases and tables

    The operating user must have RAM permissions on the corresponding cloud database resources, or have the Administrator or DBA role in DMS.

    • Method 1: RAM policy authorization

      Grant the user the RAM permission policies for the corresponding cloud service. The following table lists commonly used system policies.

      Engine type

      Action

      Resource ARN

      System policy

      RDS

      dms:LoginDatabase

      acs:rds:{region}:{accountId}:dbinstance/{instanceId}

      AliyunRDSFullAccess

      PolarDB

      dms:LoginDatabase

      acs:polardb:{region}:{accountId}:*/{instanceId}

      AliyunPolardbFullAccess

      AnalyticDB MySQL

      adb:*

      acs:adb:{region}:{accountId}:dbcluster/{instanceId}

      AliyunADBFullAccess

      AnalyticDB PostgreSQL

      gpdb:*

      acs:gpdb:{region}:{accountId}:dbinstance/{instanceId}

      AliyunGPDBFullAccess

    • Method 2: Grant the DMS Administrator or DBA role to the user

      Grant the Administrator or DBA role to the user in DMS. For more information, see .

    Memory management

    Long-term memory is a core capability of the agent. During data analysis conversations, the system automatically captures and distills key information that is valuable for future analysis, and stores it as long-term memory. These memories are intelligently recalled in subsequent interactions, improving the accuracy of the agent's understanding of your business requirements and optimizing the overall analysis experience.

    Configuration

    This feature is enabled by default to continuously optimize your analysis experience. To disable the automatic memory storage, perform the following steps:

    List description

    On the Memory page, you can view and manage all stored memories. The list contains the following information:

    Column

    Description

    Source

    Indicates the specific conversation from which the memory was generated. Click the link to go to the original conversation context and trace the source of the memory.

    Content

    Displays the key information extracted and interpreted by the system from the conversation.

    Heat

    The system calculates a popularity score based on the recall and usage frequency of each memory. A higher score indicates that the memory is more frequently used and more important in your analysis work.

    Status

    A memory can be in one of two states:
    - Remembered: The memory is active and will be recalled during subsequent analysis.
    - Forgotten: The memory is inactive and will no longer be recalled.


    Operation

    In the Actions column, you can edit and delete memories:

    • Edit: If a stored memory is inaccurate, incomplete, or if the business definition has changed, click the edit icon next to the memory entry. Manually correct or refine the memory content in the dialog box that appears, and then save the changes. This ensures that the agent learns the most accurate information.

    • Delete: If a memory is outdated, invalid, or should no longer be used in analysis, click the delete icon next to the memory entry. The behavior depends on the current state of the memory:

      • For a Remembered memory: The status changes to Forgotten, but the record is retained in the list. You can choose to remember it again.

      • For a Forgotten memory: The record is permanently removed from the list and cannot be recovered.

    Data privacy statement

    • The agent does not use your personal data, including uploaded files, query content, analysis results, and generated reports, for model training or iteration.

    • Your data is processed in dedicated compute instances that are launched on demand under your account. These instances are destroyed after release. Other accounts cannot access the data, processes, or history under your account through any means.

    • For locally uploaded data, the original files are stored in an isolated environment that is separate from other accounts and invisible to them.

    • For database-based analysis, the data remains in the database that you purchased. No data migration occurs during computation. After computation is complete, all intermediate results are destroyed and not stored.