Skip to directory

Google BigQuery Remote MCP

Google Cloud

Streamable HTTP6 documented tools14

About Google BigQuery Remote MCP

Google-managed remote MCP server for listing BigQuery datasets and tables, retrieving metadata, and running SQL queries against data the authenticated Google Cloud identity can access.

What you can use it for

  • List datasets and tables accessible to a Google Cloud identity.
  • Inspect dataset and table metadata.
  • Run read-only SQL analysis on authorized BigQuery data.
  • Run SQL that may mutate data only when explicitly requested and appropriately authorized.
Connection and setup
Hosting
Remote, Google-managed
Transport
Streamable HTTP
Endpoint
https://bigquery.googleapis.com/mcp
Authentication
OAuth 2.0 with Google Cloud IAM. The documented scope is https://www.googleapis.com/auth/bigquery; tool access also requires appropriate IAM permissions such as mcp.tools.call, bigquery.jobs.create, and bigquery.tables.getData.

Requirements

  • A Google Cloud account and selected or created Google Cloud project.
  • BigQuery API enabled for the project; new projects may enable it automatically.
  • IAM permissions for MCP tool calls, query jobs, and the data being queried.
  • A supported remote MCP client and a browser or compatible credential flow for OAuth.
  • Billing is optional for the documented sandbox path; usage of underlying BigQuery APIs remains subject to their quotas.

Setup steps

  1. Select or create a Google Cloud project and enable the BigQuery API if needed.
  2. Add the remote endpoint https://bigquery.googleapis.com/mcp to a client supporting remote MCP Streamable HTTP.
  3. Authenticate with Google Cloud OAuth and grant the BigQuery scope and required IAM permissions.
  4. Prefer `execute_sql_readonly` for analysis and inspect tool schemas before sending queries.
  5. Verify by listing datasets or reading table metadata; do not run a query that mutates data as a connection check.
Tools 6

Google Cloud official BigQuery MCP reference lists exactly six tools, matching stored names. execute_sql can mutate; execute_sql_readonly accepts read-only SQL.

  • list_dataset_ids

    List BigQuery datasets.

    01
  • list_table_ids

    List tables in a dataset.

    02
  • get_dataset_info

    Retrieve dataset metadata.

    03
  • get_table_info

    Retrieve table metadata.

    04
  • execute_sql

    Execute SQL; this is the only documented non-read-only tool.

    05
  • execute_sql_readonly

    Execute read-only SQL; does not permit DML, DDL, Python UDFs, or GQL.

    06
Limitations and notes
  • Query results are limited to 3,000 rows.
  • Queries default to a three-minute processing limit and are canceled when it is exceeded.
  • SQL tools do not support Google Drive external tables.
  • `execute_sql_readonly` excludes mutations, Python UDFs, and GQL.
  • The server itself has no separate call quota, but the underlying Google APIs enforce quotas.
  • The endpoint shown is the global endpoint; toolset-specific endpoints may be used for more granular configurations.
Source discrepancies
  • Do not conflate this Google-managed remote server with Google's separately documented local BigQuery MCP server; this record describes the remote service.
Original sources
  • official tool reference

    Google Cloud reference enumerates the six stored tools for the global endpoint https://bigquery.googleapis.com/mcp. tools/list is unauthenticated for discovery; calling operations requires cloud auth/permissions. execute_sql is read-write, while execute_sql_readonly is restricted to SELECT.

  • official documentation or source

    Google Cloud reference enumerates the six stored tools for the global endpoint https://bigquery.googleapis.com/mcp. tools/list is unauthenticated for discovery; calling operations requires cloud auth/permissions. execute_sql is read-write, while execute_sql_readonly is restricted to SELECT.

Documentation checked 2026-10-28T13:00:03.000Z. Connection has not been runtime tested.

Similar tools