MCP Tabular Data Analysis Server

You can run the MCP Tabular Data Analysis Server to analyze numeric and tabular data from CSV files and SQLite databases. It exposes practical tools for exploring datasets, quality checks, and a variety of analytics, making it easy to derive insights from tabular data with minimal setup.

How to use

You interact with the server through an MCP client to perform data analysis tasks. Start the server using your preferred runtime, then issue tool calls to inspect datasets, compute statistics, detect anomalies, and generate visualizations. You can work with CSV files or SQLite databases stored on your system and save or export results as needed.

How to install

Prerequisites include Python 3.10 or newer and a runtime like uv (recommended) or pip to install and run the server.

Install with uv

cd mcp-tabular
uv sync

Install with pip

Install the package in editable mode to run the MCP server from source.

cd mcp-tabular
pip install -e .

Run the server directly after installation.

# With uv
uv run mcp-tabular

# With pip installation
mcp-tabular

Configuration and usage notes

Configure your MCP client by pointing it to the server runtime you use. A typical local setup involves starting the Python-based server via a virtual environment and connecting to it from your MCP client.

Configuration for Claude Desktop (example integration)

If you use Claude Desktop for quick testing and prompts, add a configuration block that points to the local server runtime.

{
  "mcpServers": {
    "tabular_data": {
      "command": "/Users/kirondeb/mcp-tabular/.venv/bin/python",
      "args": [
        "-m",
        "mcp_tabular.server"
      ]
    }
  }
}

Path resolution and data samples

Paths resolve relative to the project root. You can use relative paths such as data/sample_sales.csv or an absolute path for your data files. Sample data provided includes data/sample_sales.csv and a SQLite database data/sample.db to help you test the server features.

Sample data and how to test

To create or load sample data for testing, you can rely on the included data files or run the provided script to generate a sample database.

Advanced usage notes

You can use the server to list files, describe datasets, compute correlations, filter data, group and aggregate, run SQL against SQLite databases, and list tables. Additional analytics include pivot tables, data quality reports, time series analysis, charts, merges, statistical tests, and auto insights. Export results to new CSV files as needed.

Available tools

list_data_files

List available CSV and SQLite files in the data directory with metadata.

describe_dataset

Generate statistics for a dataset including shape, types, distributions, and missing values.

detect_anomalies

Identify outliers in numeric columns using Z-score or IQR methods.

compute_correlation

Compute correlation matrices between numeric columns.

filter_rows

Filter data using operators such as eq, ne, gt, gte, lt, lte, contains, startswith, endswith.

group_aggregate

Group data and compute aggregations like sum, mean, and count.

query_sqlite

Execute SQL queries on SQLite databases.

list_tables

List all tables and schemas in a SQLite database.

create_pivot_table

Create Excel-style pivot tables with flexible aggregations.

data_quality_report

Assess data quality with scores and recommendations.

analyze_time_series

Analyze trends, seasonality, and moving averages in time series data.

generate_chart

Create visualizations such as bar, line, scatter, histogram, pie, and box plots.

merge_datasets

Join/merge two datasets with inner, left, right, or outer joins.

statistical_test

Perform hypothesis tests like t-test, ANOVA, chi-squared, and correlation tests.

auto_insights

Automatically discover patterns and insights across data.

export_data

Export filtered or transformed data to new CSV files.