MCP Server SQL Inspector

A Model Context Protocol (MCP) server that enables AI to inspect SQL databases and analyze query performance.

Tools

• list_tables: List tables in the database.
• get_schema: Get schema definition for one or more tables.
  • table_names (list): List of table names to inspect.
• get_table_stats: Get estimated row count for a table.
• run_explain: Analyze a query execution plan.
  • query: The SQL query to analyze (SELECT only).
  • just_explain_plan (optional): If true, gets the plan without running the query (default: False).

Installation

Using uv (Recommended)

When using uv no installation is necessary. You can run the server directly:

uvx mcp-server-sql-inspector

Configuration with Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "sql-inspector": {
      "command": "uvx",
      "args": ["mcp-server-sql-inspector"],
      "env": {
        "DATABASE_URL": "postgresql://user:password@localhost:5432/dbname"
      }
    }
  }
}

Local Development Configuration

If you are developing the server locally and want to test changes in Claude Desktop:

{
  "mcpServers": {
    "sql-inspector": {
      "command": "/absolute/path/to/mcp-server-sql-inspector/.venv/bin/mcp-server-sql-inspector",
      "args": [],
      "env": {
        "DATABASE_URL": "sqlite:///demo.db"
      }
    }
  }
}

From Source

git clone <repository-url>
cd mcp-server-sql-inspector
poetry install --extras all

Docker

Build the image:

docker build -t mcp-server-sql-inspector .

Run the container:

docker run -i --rm -e DATABASE_URL="postgresql://user:password@host:port/dbname" mcp-server-sql-inspector

Docker Configuration for Claude Desktop

{
  "mcpServers": {
    "sql-inspector": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e", "DATABASE_URL=postgresql://user:password@host:port/dbname",
        "mcp-server-sql-inspector"
      ]
    }
  }
}

Configuration

The server requires a DATABASE_URL environment variable to connect to your database.

Supported database protocols (requires appropriate drivers):

• PostgreSQL: postgresql://...
• MySQL: mysql://...
• MSSQL: mssql+pyodbc://...
• SQLite: sqlite:///...

Development

1. Install dependencies:

   poetry install --extras all
   

2. Run the server:

   export DATABASE_URL="sqlite:///test.db"
   poetry run mcp-server-sql-inspector
   

License

MIT