Skip to content

README.md

Latest commit

 

History

History
74 lines (50 loc) · 2.97 KB

README.md

File metadata and controls

74 lines (50 loc) · 2.97 KB
Raw

Google Cloud BigQuery

Cursor plugin that connects agents to Google Cloud BigQuery through Google Cloud BigQuery's official remote Model Context Protocol server.

Explore datasets and tables and run SQL queries.

Install

  1. Open Cursor Settings → Plugins.
  2. Search for Google Cloud BigQuery.
  3. Click Install, then complete the Google Cloud BigQuery sign-in prompt.

Or run /add-plugin google-cloud-bigquery in chat.

MCP

{
  "mcpServers": {
    "google-cloud-bigquery": {
      "type": "http",
      "url": "https://bigquery.googleapis.com/mcp",
      "auth": {
        "CLIENT_ID": "${CLIENT_ID}",
        "CLIENT_SECRET": "${CLIENT_SECRET}"
      }
    }
  }
}

Auth is OAuth. Cursor prompts for Google Cloud BigQuery sign-in when the plugin connects. It signs in with a Google OAuth client your team provides; see below.

Set up your team's OAuth client

A team admin provides one Google OAuth client, and every member signs in under it, so consent and audit show the team's own app. The Client ID and Client secret are required.

  1. In the Google Cloud console, create an OAuth client for a Web application and add Cursor's OAuth callback URL as an authorized redirect URI. Add the https://www.googleapis.com/auth/bigquery scope to the project's OAuth consent screen.
  2. In the Cursor dashboard, open the team's plugin settings for Google Cloud BigQuery and enter the Client ID and Client secret.

Members then sign in to BigQuery under that client. The secret is stored encrypted and stays on Cursor's backend; it is never sent to members' machines. Changing the client signs members out, and they reconnect once.

Before you connect

You need a Google Cloud project with the BigQuery API enabled, and your Google account needs the MCP Tool User role (roles/mcp.toolUser) plus BigQuery Job User and Data Viewer on that project.

What agents can do

Category Capabilities
Explore List datasets and tables and read their metadata and schemas
Query Run read-only SQL, including BigQuery ML functions
Write Run DML and DDL when the account allows it
Jobs Poll results and cancel running query jobs

The hosted runtime is the source of truth for tool names and schemas.

Notes

  • Queries are billed to the project id passed to the tool.
  • Admins can block the write-capable execute_sql tool with an IAM deny policy while leaving read-only queries available.
  • Grok also offers this connector with service-identity auth; this plugin configures only the user OAuth flow.

Docs

Logo is the official BigQuery product icon from Google Cloud's architecture icon set, on a padded white tile.

License

MIT