SQLite MCP Server

An MCP (Model Context Protocol) server that enables LLMs to interact with SQLite databases. This server provides read-only access to any SQLite database through a set of tools for schema inspection and SQL query execution.

Built with async/await patterns using aiosqlite for non-blocking database operations.

Features

• list_tables - Get all table names in the database
• describe_table - Get column details, types, and relationships for a specific table
• get_schema - Get complete database schema as CREATE TABLE statements
• query - Execute read-only SQL queries (SELECT only)

Prerequisites

• Python 3.13+
• uv package manager

Installation

1. Clone this repository
2. Install dependencies:

uv sync

Quick Start with Sample Data

This project includes a CRM sample database for demonstration purposes.

1. Create and seed the database

uv run seed_database.py

This creates crm.db with sample CRM data:

• 5 users (sales team)
• 20 companies
• 40 contacts
• 25 deals
• 30 activities

2. Run the MCP server

See Running the Server below.

Running the Server

The server uses crm.db in the same directory as the script (hardcoded).

Development Mode (MCP Inspector)

uv run mcp dev sqlite_mcp_server.py

This launches the MCP Inspector - a web-based debugging UI at http://localhost:5173 where you can:

• See all registered tools
• Test each tool interactively with custom inputs
• Inspect JSON responses
• Debug issues before connecting an LLM

Production Mode (LLM Integration)

uv run sqlite_mcp_server.py

This starts the server in stdio mode, ready for LLM clients to connect.

Client Configuration

Example: Claude Desktop

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "sqlite": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/sqlite-mcp-server",
        "run",
        "sqlite_mcp_server.py"
      ]
    }
  }
}

Tool Reference

list_tables

Returns all table names in the database.

Example response:

{
  "tables": ["activities", "companies", "contacts", "deals", "users"]
}

describe_table

Get detailed information about a table's structure.

Parameters:

• table_name (string): Name of the table to describe

Example response:

{
  "table": "contacts",
  "columns": [
    {"name": "id", "type": "INTEGER", "nullable": false, "primary_key": true},
    {"name": "first_name", "type": "TEXT", "nullable": false, "primary_key": false},
    {"name": "email", "type": "TEXT", "nullable": true, "primary_key": false}
  ],
  "foreign_keys": [
    {"column": "company_id", "references_table": "companies", "references_column": "id"}
  ],
  "indexes": []
}

get_schema

Returns complete database schema as CREATE TABLE statements.

Example response:

{
  "schema": {
    "users": "CREATE TABLE users (...)",
    "companies": "CREATE TABLE companies (...)"
  }
}

query

Execute a read-only SQL query. Only SELECT statements are allowed.

Parameters:

• sql (string): A SELECT SQL statement

Example response:

{
  "columns": ["name", "email"],
  "rows": [
    ["John Doe", "john@example.com"],
    ["Jane Smith", "jane@example.com"]
  ],
  "row_count": 2
}

Security: The server validates that queries are read-only by:

1. Ensuring queries start with SELECT or WITH
2. Blocking keywords like INSERT, UPDATE, DELETE, DROP, etc.

Using Your Own Database

The server is hardcoded to use crm.db in the script directory. To use a different database:

1. Replace crm.db with your database file (keep the same name), or
2. Modify the DB_PATH constant in sqlite_mcp_server.py