MCP Server for Spanner

What is MCP Server for Spanner?

MCP Server for Spanner is a Model Context Protocol (MCP) server that connects to Google Cloud Spanner databases and executes SQL queries. It is designed for AI assistants to interact with Spanner databases.

How to use MCP Server for Spanner?

Clone the repository, install dependencies (npm install), build (npm run build), then configure the server using environment variables (PROJECT_ID, INSTANCE_ID, DATABASE_ID, API_ENDPOINT) or command-line arguments. The server can be launched via Claude Desktop settings or with MCP Inspector for development.

Key features of MCP Server for Spanner

• Execute SELECT queries on Spanner tables
• Run INSERT, UPDATE, DELETE data modification queries
• Execute DDL statements (CREATE TABLE, ALTER TABLE, etc.)
• List all tables in the database
• Show detailed table structure (schema)

Use cases of MCP Server for Spanner

• AI assistants querying Spanner data via natural language
• Automated schema exploration and documentation
• Developing and testing Spanner applications with an emulator
• Performing ad‑hoc operations on a Spanner emulator during development

FAQ from MCP Server for Spanner

What is MCP Server for Spanner?

It is an MCP server that enables AI assistants to interact with Google Cloud Spanner databases by executing SQL queries and retrieving schema information.

What are the runtime requirements?

Node.js 22 or later is required. The tool currently only supports connecting to a Spanner emulator (not a production Spanner instance).

What data can be accessed?

The server can read and modify any data in the configured Spanner database, including tables, rows, and schema objects. It does not access external systems.

How does the server communicate?

It uses the standard MCP transport over STDIO, as configured in the MCP client (e.g., Claude Desktop).

How is authentication handled?

Authentication is configured via environment variables (PROJECT_ID, INSTANCE_ID, DATABASE_ID, API_ENDPOINT). For the emulator, no actual Google Cloud credentials are needed; the emulator endpoint is specified directly.