bigquery-mcp
About
An MCP server to help AI Agents inspect the contents of a BigQuery warehouse
Details
- Author
- PaddyAlton
- GitHub stars
- 1
- Downloads
- 388
- Categories
- Database, AI
Jump to
- Provides AI agents with tools to list BigQuery datasets, tables, and columns.
- Lets agents inspect query history for context.
- Helps agents write better SQL by exposing database contents.
- Easy integration with Cursor IDE via a single command.
- Relies on uv for dependency management and BigQuery client library for authentication.
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
bigquery-mcpCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
After ensuring uv is installed and a BigQuery client is available (via gcloud or a service account key set in GOOGLE_APPLICATION_CREDENTIALS), clone the repository and run uv sync. Then, in Cursor settings under MCP Servers, start a server with the command: uv run --with mcp --directory /path/to/bigquery-mcp mcp run /path/to/bigquery-mcp/src/server.py. A contextual rule file (.cursor/rules/tool-use-rule.mdc) is recommended to give the Cursor Agent proper instructions for using the tools.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"bigquery-mcp": {
"bigquery-mcp": {
"command": "uv",
"args": [
"run",
"--with",
"mcp",
"--directory",
"/path/to/bigquery-mcp",
"mcp",
"run",
"/path/to/bigquery-mcp/src/server.py"
],
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": ""
}
}
}
}
}McpServers
{
"bigquery-mcp": {
"command": "uv",
"args": [
"run",
"--with",
"mcp",
"--directory",
"/path/to/bigquery-mcp",
"mcp",
"run",
"/path/to/bigquery-mcp/src/server.py"
],
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": ""
}
}
}bigquery-mcp
A Model Context Protocol (MCP) Server for BigQuery.
Prerequisites
This project and the mcp CLI rely on your having the dependency management tool uv installed. You can install via e.g. brew install uv for Homebrew users. See here for alternatives.
This project (currently) assumes you can 'transparently' create a BigQuery Client, which is usually the case if you have gcloud installed in your local environment. In other environments you may need suitable service account credentials (and you can set the GOOGLE_APPLICATION_CREDENTIALS environment variable to indicate the location of these credentials to the BigQuery client library).
For development
You will also need to have Taskfile installed. brew install go-task will work if you are a Homebrew user.
See here for alternatives.
Quickstart for Cursor IDE
0. ensure you have the prerequisites installed
1. clone down this repository
2. run uv sync to install the dependencies
3. in Cursor settings > MCP Servers, start a server with the following command:
uv run --with mcp --directory /path/to/bigquery-mcp mcp run /path/to/bigquery-mcp/src/server.py
It is recommended to write a contextual rule in .cursor/rules/tool-use-rule.mdc into your working project. The Cursor Agent will need some instructions if it is
to use the tools properly.
Details
AI Agents have begun to excel at writing code, but often struggle with data-related tasks. This is because of the coupling between programme logic and the actual contents of the database.
More specifically, AI Agents often fail to write good SQL queries for analysis tasks. They are capable of writing code, so the issue is not a lack of ability in this arena and more due to a lack of context about the _contents_ of the database.
This MCP server assists with this problem area by providing AI Agents with tools they can use to examine the contents of a BigQuery data warehouse (i.e. datasets, tables, columns, query history).
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
