BigQuery MCP Server for Claude

Content

The BigQuery MCP server connects Claude and other MCP-capable clients to BigQuery through Google's managed remote HTTP MCP endpoint at https://bigquery.googleapis.com/mcp. It is built for warehouse analytics workflows where an assistant needs to discover datasets and tables, inspect metadata, run read-only SQL, and summarize results without building a custom database connector.

This entry defaults to a conservative usage model: grant only the Google Cloud IAM permissions needed for MCP tool calls, BigQuery job creation, and data viewing, prefer execute_sql_readonly, and use Google Cloud IAM deny policies when the write-capable execute_sql tool should not be available. BigQuery MCP traffic is authenticated with OAuth 2.0 and IAM rather than API keys.

Features

• Remote HTTP MCP endpoint at https://bigquery.googleapis.com/mcp.
• OAuth 2.0 and Google Cloud IAM based authentication and authorization.
• Dataset and table discovery with list_dataset_ids and list_table_ids.
• Dataset and table metadata lookup with get_dataset_info and get_table_info.
• Read-only SQL execution through execute_sql_readonly, restricted to SELECT statements.
• General SQL execution through execute_sql when IAM and policy allow it.
• Automatic goog-mcp-server:true query job label for read-only SQL jobs.
• Google Cloud security controls, including IAM deny policies and optional Model Armor floor settings for Google MCP server traffic.
• BigQuery query limits documented for MCP use, including three-minute default query processing time and 3,000-row result limits.

Use Cases

• Ask Claude to list datasets and identify candidate tables for an analytics question in a specific Google Cloud project.
• Inspect schemas and metadata before writing a warehouse query.
• Run bounded read-only SQL to summarize sales, product, support, or operations metrics.
• Find BigQuery jobs run through the MCP server by filtering for the goog-mcp-server:true job label.
• Use BigQuery forecasting or other advanced BigQuery capabilities after human review of the generated SQL and cost impact.
• Build agent workflows that use approved BigQuery insights to draft tickets, reports, emails, or follow-up analysis.

Installation

Claude Code

1. Confirm the BigQuery API is enabled for the target Google Cloud project.
2. Ask an administrator to grant the user or agent identity the required IAM permissions for MCP tool calls, BigQuery jobs, and BigQuery data viewing.
3. Add the remote MCP server:

claude mcp add --transport http bigquery https://bigquery.googleapis.com/mcp

1. Complete the Google OAuth or client-specific authentication flow.
2. Start with narrow read-only prompts that name the project, dataset, table, region, and expected result shape.

Claude Desktop

1. Open the Claude Desktop MCP configuration file.
2. Add the bigquery HTTP server configuration shown below.
3. Restart Claude Desktop and complete the Google authentication flow when the client prompts for it.
4. Test by listing datasets in a non-production or sandbox project first.

Configuration

{
  "mcpServers": {
    "bigquery": {
      "type": "http",
      "url": "https://bigquery.googleapis.com/mcp"
    }
  }
}

Examples

List datasets in a project

List the BigQuery datasets in project PROJECT_ID and summarize what each one appears to contain.

Inspect table schema

For project PROJECT_ID and dataset DATASET_ID, list tables and inspect metadata for the likely orders table.

Run a bounded read-only query

Use read-only SQL to show the top 10 orders by volume from PROJECT_ID.DATASET_ID.TABLE_ID.

Review MCP query jobs

Find recent BigQuery jobs in project PROJECT_ID with the label goog-mcp-server:true and summarize their query purpose.

Source notes

• The official Google Cloud guide describes the BigQuery remote MCP server for connecting AI applications including Claude, ChatGPT, Gemini CLI, and custom clients to BigQuery for running queries, getting metadata, and listing resources.
• Google documents the endpoint as https://bigquery.googleapis.com/mcp with HTTP transport, OAuth 2.0, IAM authorization, and no API-key support.
• The guide lists required roles for MCP tool calls, BigQuery job creation, and BigQuery data viewing, plus required permissions such as mcp.tools.call, bigquery.jobs.create, and bigquery.tables.getData.
• The MCP reference lists the BigQuery tools: list_dataset_ids, get_dataset_info, list_table_ids, get_table_info, execute_sql_readonly, and execute_sql.
• Google documents that execute_sql_readonly is SELECT-only, while execute_sql can run broader BigQuery SQL and should be restricted when read-only use is intended.

Duplicate check

Checked current content/mcp/, content/tools/, guides, skills, agents, open pull requests, live HeyClaude llms-full.txt, and repository-wide content for BigQuery MCP, bigquery.googleapis.com/mcp, Google Cloud MCP, execute_sql_readonly, mcp.toolUser, warehouse analytics, Snowflake, and data warehouse. No dedicated BigQuery MCP entry, BigQuery MCP endpoint source URL duplicate, or open duplicate PR was found.

Disclosure

Editorial listing. No paid placement or affiliate link is used.

• Source repositorygithub.com 2026-07-18T19:14:44+00:00
• Documentationdocs.cloud.google.com
• Submitted by oktofeesh12026-06-03