csv-postgres-loader 0.1.0

MCP server to load CSV files to PostgreSQL with validation and progress tracking

CSV to PostgreSQL MCP Server

An MCP (Model Context Protocol) server that loads CSV files into PostgreSQL databases with validation and progress tracking.

Features

• CSV Validation: Validates CSV structure and provides detailed error messages
• Efficient Loading: Uses PostgreSQL COPY command for fast bulk loading
• Progress Tracking: Shows progress bar for long-running imports
• Flexible Configuration: Optional database name (defaults to csvimports)
• Error Reporting: Exact line numbers and column information for validation errors

Installation

From PyPI

pip install csv-postgres-loader

Or using uv:

uv pip install csv-postgres-loader

From Source

Clone the repository and install with uv:

git clone https://github.com/raviramadoss/mcp-csv-postgres.git
cd mcp-csv-postgres
uv sync --all-extras

Usage

As MCP Server

Option 1: After installing from PyPI

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

{
  "mcpServers": {
    "csv-postgres-loader": {
      "command": "uv",
      "args": [
        "tool",
        "run",
        "--from",
        "csv-postgres-loader",
        "csv-postgres-loader"
      ]
    }
  }
}

Option 2: For local development

{
  "mcpServers": {
    "csv-postgres-loader": {
      "command": "uv",
      "args": [
        "tool",
        "run",
        "--from",
        "/path/to/mcp-csv-postgres",
        "--python",
        "3.10",
        "csv-postgres-loader"
      ]
    }
  }
}

Or run directly from command line:

csv-postgres-loader

Tool: load_csv_to_postgres

Loads a CSV file into PostgreSQL database.

Parameters:

• file_path (required): Path to the CSV file
• dbname (optional): Database name (default: csvimports)
• host (optional): PostgreSQL host (default: localhost)
• port (optional): PostgreSQL port (default: 5432)
• user (optional): PostgreSQL user (default: postgres)
• password (optional): PostgreSQL password
• table_name (optional): Table name (derived from filename if not provided)

Returns:

• Success message with database name, table name, and rows loaded
• Error message with detailed validation or database errors

Testing the Server

Quick Test

Run the test script to see the server in action:

uv run python test_server.py

This will:

1. Create sample CSV files
2. Test loading them into PostgreSQL
3. Test validation error reporting
4. Show detailed output

Manual Testing with Sample Data

A sample CSV file is provided in sample_data.csv. To test manually:

# In Python/IPython
import asyncio
from csv_postgres_loader.server import call_tool

result = asyncio.run(call_tool(
    "load_csv_to_postgres",
    {"file_path": "sample_data.csv"}
))
print(result[0].text)

Using with MCP Inspector

To test with the MCP Inspector tool:

npx @modelcontextprotocol/inspector uv run csv-postgres-loader

VS Code Launch Configurations

The .vscode/launch.json includes:

• Run MCP Server: Start the server in debug mode
• Test MCP Server with Sample CSV: Run the test script
• Run All Tests: Execute full test suite
• Run Specific Test: Debug a single test file

Development

Running Tests

uv run pytest

Code Coverage

uv run pytest --cov-report=html

Coverage report will be available in htmlcov/index.html.

Current coverage: >90%

Pre-commit Hooks

Pre-commit hooks are configured to run tests before each commit:

uv run pre-commit install
uv run pre-commit run --all-files

Architecture

Modules

• validator.py: CSV validation with detailed error reporting

  • File existence and readability checks
  • CSV structure validation
  • Row consistency validation
  • Automatic dialect detection

• database.py: PostgreSQL database operations

  • Connection management
  • Database creation
  • Table creation from CSV headers
  • Row counting

• loader.py: CSV loading using COPY command

  • Progress tracking with tqdm
  • Efficient bulk loading
  • Error handling and rollback

• server.py: MCP server implementation

  • stdio transport
  • Tool registration
  • Error handling

Error Handling

The server provides detailed error messages:

• Validation Errors: Exact line numbers and description of CSV issues
• Database Errors: Connection, permission, and SQL errors
• File Errors: Permission, encoding, and file not found errors

Examples

Loading a CSV file

{
  "tool": "load_csv_to_postgres",
  "arguments": {
    "file_path": "/path/to/data.csv"
  }
}

With custom database

{
  "tool": "load_csv_to_postgres",
  "arguments": {
    "file_path": "/path/to/data.csv",
    "dbname": "mydb",
    "table_name": "my_table"
  }
}

License

MIT