Reference resources

This page describes how to reference Google Cloud data resources when working with Google Cloud Data Agent Kit. You can reference resources in your development environment or supported coding agent using explicit or implicit referencing, depending on the environment and agent that you use.

  • Explicit referencing. You explicitly identify a resource in chat or the user interface, such as by typing @RESOURCE_NAME, selecting a resource from the Model Context Protocol (MCP) resource menu, or using drag-and-drop.
  • Implicit referencing. You refer to a resource through natural language context in chat, such as typing "this table" or "this dataset", referring to the open or selected resource.

Referencing methods

If your development environment and coding agent support referencing, you can reference data resources in the following ways:

  • Chat mentions: In supported chat interfaces, type @ followed by the resource name or ID (for example, @DATASET_NAME) to explicitly attach the resource context to your prompt.
  • MCP resource menu: Select a resource directly from the Model Context Protocol (MCP) resource picker in your IDE.
  • Drag-and-drop: Drag a data asset from the explorer tree directly into the chat or editor window.
  • Natural language reference: In environments that support implicit referencing, use contextual phrases like "this table" or "this dataset" to have the agent automatically infer the active resource.

The following table describes the supported referencing methods for each development environment and coding agent.

Development environment or agent Chat mention (@) MCP resource menu Drag-and-drop Natural language
VS Code (GitHub Copilot), Cursor, Antigravity IDE Supported Supported Supported Supported
Claude Code CLI (standalone) Supported Not supported Not supported Supported
Antigravity 2.0, Antigravity CLI, Claude Code (VS Code extension) Not supported Not supported Not supported Supported

Supported resources

The following table describes the referencing syntax and natural language phrases for supported Google Cloud resources across services.

  1. Partial table name is also supported.
  2. In Claude Code CLI, use @SESSION_ID.
  3. In Claude Code CLI, use @BATCH_ID. Implicit referencing is not supported.
Service Resource Chat mention syntax (@) Natural language reference
BigQuery Dataset1 @DATASET_NAME "this dataset"
  Table @TABLE_NAME "this table"
  View @VIEW_NAME "this view"
AlloyDB for PostgreSQL, Cloud SQL, Spanner Table @TABLE_NAME "this table"
  View @VIEW_NAME "this view"
Spark Cluster @CLUSTER_NAME "this cluster"
  Serverless runtime @RUNTIME_NAME "this runtime"
  Job @JOB_NAME "this job"
  Interactive session2 @SESSION_NAME "this session"
  Batch3 @BATCH_NAME Not supported

What's next