Home / MCP / MCP BigQuery Server

MCP BigQuery Server

Provides MCP access to Google BigQuery datasets via a lightweight Go server.

Installation

Add the following to your MCP client configuration file.

Configuration

{
    "mcpServers": {
        "bigquery": {
            "command": "mcp-bigquery-server",
            "args": [
                "start",
                "--project",
                "sample-project",
                "--dataset",
                "test1",
                "--dataset",
                "test2"
            ]
        }
    }
}

MCP BigQuery Server lets you query BigQuery datasets through MCP, providing a lightweight Go-based endpoint you can start locally and connect to with an MCP client to run dataset, table, and schema queries plus dry-run cost estimates.

How to use

You will run the MCP BigQuery Server locally and point your MCP client to it as an MCP server. Start the server with your project and one or more BigQuery datasets, then perform actions from your client such as listing available datasets, listing tables in a dataset, inspecting table schemas, and performing a dry-run of a query to estimate cost and time.

How to install

Prerequisites: macOS with Homebrew installed. You will use Homebrew to install the MCP BigQuery Server binary, then start it from the command line.

Install via Homebrew:

$ brew tap monochromegane/tap
$ brew install monochromegane/tap/mcp-bigquery-server

Registration and configuration

Configure your MCP client to use the BigQuery server by registering it under the mcpServers section. You specify the command to run and any arguments needed to start the server. You can list multiple datasets by repeating the --dataset argument.

{
  "mcpServers": {
    "BigQuery": {
      "command": "mcp-bigquery-server",
      "args": [
        "start",
        "--project",
        "sample-project",
        "--dataset",
        "test1",
        "--dataset",
        "test2"
      ]
    }
  }
}

Notes about configuration

You can specify multiple datasets by repeating the --dataset argument. Ensure each dataset you reference is valid within your BigQuery project to avoid startup or query errors.

Available tools

The server exposes a set of actions you can perform from your MCP client. Typical actions include fetching datasets, listing tables, retrieving table schemas, and performing a dry run of a query to estimate cost and duration.

Available tools

list_allowed_datasets

Returns a list of all datasets that the server is configured to allow access to.

list_tables

Provides a detailed listing of all tables within a specified dataset.

get_table_schema

Retrieves the schema for a specified table in a specified dataset.

dry_run_query

Performs a dry run of a query to estimate cost and execution time without running the query.