genai-toolbox/docs/CLOUDSQLMSSQL_README.md

Cloud SQL for SQL Server MCP Server

The Cloud SQL for SQL Server Model Context Protocol (MCP) Server gives AI-powered development tools the ability to work with your Google Cloud SQL for SQL Server databases. It supports connecting to instances, exploring schemas, and running queries.

Features

An editor configured to use the Cloud SQL for SQL Server MCP server can use its AI capabilities to help you:

• Query Data - Execute SQL queries
• Explore Schema - List tables and view schema details

Installation and Setup

Prerequisites

• Download and install MCP Toolbox:

1. Download the Toolbox binary: Download the latest binary for your operating system and architecture from the storage bucket. Check the releases page for OS and CPU architecture support: https://storage.googleapis.com/genai-toolbox/v0.21.0/<os>/<arch>/toolbox

   • Replace <os> with linux, darwin (macOS), or windows.
   • Replace <arch> with amd64 (Intel) or arm64 (Apple Silicon).

   curl -L -o toolbox https://storage.googleapis.com/genai-toolbox/v0.21.0/linux/amd64/toolbox
   

2. Make it executable:

   chmod +x toolbox
   

3. Add the binary to $PATH in .~/bash_profile:

   export PATH=$PATH:/path/to/toolbox
   

Note: You may need to restart Antigravity for changes to take effect. Windows OS users will need to follow one of the Windows-specific methods.

• A Google Cloud project with the Cloud SQL Admin API enabled.
• Ensure Application Default Credentials are available in your environment.
• IAM Permissions:
  • Cloud SQL Client (roles/cloudsql.client)

Configuration

The MCP server is configured using environment variables.

export CLOUD_SQL_MSSQL_PROJECT="<your-gcp-project-id>"
export CLOUD_SQL_MSSQL_REGION="<your-cloud-sql-region>"
export CLOUD_SQL_MSSQL_INSTANCE="<your-cloud-sql-instance-id>"
export CLOUD_SQL_MSSQL_DATABASE="<your-database-name>"
export CLOUD_SQL_MSSQL_USER="<your-database-user>"  # Optional
export CLOUD_SQL_MSSQL_PASSWORD="<your-database-password>"  # Optional
export CLOUD_SQL_MSSQL_IP_TYPE="PUBLIC"  # Optional: `PUBLIC`, `PRIVATE`, `PSC`. Defaults to `PUBLIC`.

Note: If your instance uses private IPs, you must run the MCP server in the same Virtual Private Cloud (VPC) network.

Add the following configuration to your MCP client (e.g., settings.json for Gemini CLI):

{
  "mcpServers": {
    "cloud-sql-mssql": {
      "command": "toolbox",
      "args": ["--prebuilt", "cloud-sql-mssql", "--stdio"],
      "env": {
        "CLOUD_SQL_MSSQL_PROJECT": "your-project-id",
        "CLOUD_SQL_MSSQL_REGION": "your-region",
        "CLOUD_SQL_MSSQL_INSTANCE": "your-instance-id",
        "CLOUD_SQL_MSSQL_DATABASE": "your-database-name",
        "CLOUD_SQL_MSSQL_USER": "your-username",
        "CLOUD_SQL_MSSQL_PASSWORD": "your-password"
      }
    }
  }
}

Usage

Once configured, the MCP server will automatically provide Cloud SQL for SQL Server capabilities to your AI assistant. You can:

• "Select top 10 rows from the customers table."
• "List all tables in the database."

Server Capabilities

The Cloud SQL for SQL Server MCP server provides the following tools:

Tool Name	Description
execute_sql	Use this tool to execute SQL.
list_tables	Lists detailed schema information for user-created tables.

Documentation

For more information, visit the Cloud SQL for SQL Server documentation.