SQL Generative

An interactive Model Context Protocol (MCP) server that translates plain English queries into SQLAlchemy operations, queries a PostgreSQL database under strict read-only execution modes, and returns the query results formatted as structured XML.

---

🌟 Key Features

• Natural Language DB Querying: Translate user queries in plain English to corresponding database results dynamically.
• SQLAlchemy Table Reflection: Reflects database schemas dynamically across multiple namespaces using standard ORM capabilities (no direct pg_dump client tools needed).
• Enforced Read-Only Transactions: Database URL connections are injected with query options (?options=-c%20default_transaction_read_only%3Don) to natively block any modifications (CREATE, UPDATE, DELETE, DROP) at the database engine level.
• Dynamic XML Response Formatting: Converts database records into a clean, well-formed XML structure automatically compiled by Claude.
• FastMCP Protocol Integration: Built on top of the standard fastmcp SDK to run as a local stdio MCP server.

---

🏗️ Architecture

---

🛠️ Setup & Installation

Prerequisites

• Python 3.13+
• uv (fast Python package installer and resolver)
• PostgreSQL Server with an active database

1. Project Initialization & Dependencies

Initialize the project environment and install dependencies:

uv sync

2. Environment Configuration

Create a .env file in the root directory and configure the database URL along with your Anthropic API key:

# Database connection for local development (enforced read-only mode via query options)
DATABASE_URL=postgresql://postgres:admin@localhost:5432/expence_db?options=-c%20default_transaction_read_only%3Don

# Anthropic API Key
ANTHROPIC_API_KEY="your-anthropic-api-key-here"

---

🚀 Running the Server

Start the stdio-based MCP server locally:

uv run python main.py

---

🔌 Integration with Claude Desktop

To configure Claude Desktop to use this database query tool on Windows, configure your claude_desktop_config.json file.

1. Press Win + R, type %APPDATA%\Claude and press Enter.
2. Open claude_desktop_config.json and insert the following server config:

{
  "mcpServers": {
    "db-query-server": {
      "command": "uv",
      "args": [
        "--directory",
        "c:/Projects/generative-tool",
        "run",
        "python",
        "main.py"
      ]
    }
  }
}

3. Restart Claude Desktop. You will see the tools icon 🔌 in the chat input area.

---

📈 Usage Examples

Once connected, you can ask Claude queries like:

• "Show a list of all tables in the database"
• "How many expense entries are recorded for AWS?"
• "Show a list of all rows in table exp_expence_core_t"

The server returns results formatted as well-formed XML:

<data>
  <item>
    <exp_expence_core_id>26</exp_expence_core_id>
    <title>AWS Bill</title>
    <amount>100.00</amount>
  </item>
</data>