arunak1998/mcp_csv_server

MCP Polars SQL Analyst Server

High-performance Model Context Protocol (MCP) server for CSV analysis and SQL execution using Polars

---

🚀 Overview

The MCP Polars SQL Analyst Server enables AI agents and developers to interact with CSV datasets through natural language or SQL queries.
It leverages the Polars DataFrame engine (Rust-powered, multi-threaded, lazy evaluation) for lightning-fast analytics on large files.

With this server, you can:

• ✅ Discover CSV files available in a directory
• ✅ Inspect schemas, datatypes, and stats
• ✅ Run SQL queries directly on your CSVs using Polars SQL
• ✅ Easily integrate with AI agents and dashboards for interactive CSV insights

---

✨ Features

• High-Performance Data Processing

  • Rust-based Polars engine (faster than pandas)
  • Multi-threaded execution across CPUs
  • Lazy evaluation for optimized queries
  • Memory-efficient for large datasets

• MCP Tools Available

  1. get_files_list

     • 📂 Retrieve all available CSV files in the configured directory
     • Returns: List of file names and paths

  2. get_schema

     • 🔎 Extract schema: column names, datatypes, and basic stats
     • Params: file_path (CSV file path)
     • Returns: JSON with field info

  3. execute_polars_sql

     • 📝 Run SQL queries on CSV data via Polars SQL engine
     • Params: sql_query, file_path
     • Returns: Structured query results

---

🛠️ Technology Stack

• MCP Framework: FastMCP – rapid protocol server creation
• Data Engine: Polars – fast, Rust-native DataFrame library
• Query Language: SQL (via Polars SQL context)
• File Support: CSV datasets

📦 Installation bash

Clone repository

git clone https://github.com/yourusername/mcp-polars-analyst.git cd mcp-polars-analyst

Install dependencies

pip install -e . ▶️ Usage Start the MCP server and point to your CSV directory:

bash FILE_LOCATION="/mnt/c/workspaces/mcpserver/temp/*.csv"
python server/analyst.py The server will:

Scan the specified location for CSVs

Initialize Polars SQL context

Expose MCP tools (get_files_list, get_schema, execute_polars_sql)

Start SSE (Server-Sent Events) for agent communication