Agent Data Gateway

更新时间:
复制 MD 格式

Agent Data Gateway is a data security gateway provided by AI-Native Database Service. Deployed between AI agents and enterprise data sources, Agent Data Gateway provides unified identity verification, access control, data masking, and full-link auditing for agent-initiated data access. No modifications to your agent code or database configurations are required.

Background

As AI agents such as Cursor, Claude Code, Qoder, and Coze become widely adopted in enterprises, agents are increasingly accessing corporate databases directly. Traditional database connection methods lack agent-specific identity management, access control, and operation auditing capabilities, introducing risks of data leakage and unintended operations.

Agent Data Gateway serves as a security proxy layer between AI agents and enterprise data. The gateway supports nearly a hundred data source types and integrates with mainstream agent clients through standard protocols, ensuring that every data operation is traceable, controllable, and subject to circuit-breaking.

Note

Agent Data Gateway is available in the following regions: China (Beijing), China (Shanghai), China (Shenzhen), Singapore, and China (Hong Kong) .

Core capabilities

Agent Data Gateway provides four core capabilities across the dimensions of access control, semantic understanding, data protection, and stability assurance:

Dimension

Capability

Description

Access

Agent identity management

Assigns a unique identity and access token to each AI agent to distinguish different agent access sources. Database connection credentials are physically isolated within the gateway and cannot be obtained by agents.

Permissions and rules

Configures fine-grained data access permissions based on agent identity, down to the database, table, column, and row levels, to control the scope of data that an agent can access.

Understanding

DataWiki semantic layer

Accumulates field semantics, metric definitions, and business rules to provide data context for agents. This enables agents to understand data meanings and business intent before querying, which improves the accuracy of natural language to SQL conversion.

Protection

Masking and constraints

Automatically detects sensitive data in query results, such as phone numbers, ID card numbers, and bank card numbers, and applies masking by default. Unauthorized access requests are blocked by default.

Full-link auditing

Records complete details of every data operation, including agent identity, SQL statements, operator, execution status, affected rows, and execution duration, to meet enterprise compliance and auditing requirements.

Stability

Emergency circuit-breaking

Cuts off an agent's data access channel within seconds when abnormal operations or risky behavior is detected. Circuit-breaking granularity can be set to a specific agent, a specific data source, or a specific operation type.

Access Agent Data Gateway

  1. Log on to the AI-Native Database Service console.

  2. In the left-side navigation pane, click Data Gateway to go to the Agent Data Gateway console.

Console navigation

Agent Data Gateway has its own console. The left-side navigation pane provides the following modules:

Module

Description

Overview

The home page of Agent Data Gateway. Displays the product introduction, administrator guide, and quick start steps.

Data Sources

Register and manage the database instances to be governed by the gateway.

Agent Identity

Create and manage access tokens, and complete agent onboarding configuration through the CLI.

My Permissions

View the resource access permissions granted to the current user.

User Management

Manage users and roles, and configure permission assignments.

Audit and Observability

View SQL execution records. Filter audit logs by operator, database, and time range.

Supported agent clients and integration methods

  • Supported agent clients: Cursor, Qoder, Claude Code, Coze, Data Agent, and custom enterprise agents.

  • Supported integration methods:

    Integration method

    Description

    Skill / OpenAPI

    RESTful API integration, suitable for custom enterprise agents.

    CLI

    Command-line tool integration, suitable for scripting and automation scenarios.

    IDE plugin

    Native integration with IDEs such as Cursor, VS Code, and DataGrip.

    Web / IM

    Browser-based access and integration with instant messaging platforms such as DingTalk and WeCom.

Data privacy and security

Security commitment

Description

Credential isolation

Database usernames and passwords are physically isolated within the gateway. Agents cannot obtain these credentials.

Default sensitive data masking

Personally identifiable information (PII) fields such as phone numbers, ID card numbers, and bank card numbers are automatically detected and masked by default. No additional configuration is required.

Default unauthorized access blocking

Access requests that cross permission boundaries, sensitivity levels, or environments are blocked by default.

Full-link traceability

Every access event is logged with complete trace information, including agent identity, responsible entity, SQL statements, triggered rules, and execution results.

Compliance auditing

Meets enterprise internal audit and regulatory compliance requirements.