CSV Analytics MCP Server

CSV Analytics MCP Server

By MEOK AI Labs | meok.ai

Spreadsheet and CSV analysis toolkit for AI agents. Load CSV files, filter and query data, compute statistics, create aggregations, build pivot tables, and export chart-ready data -- all powered by pandas.

Tools

Installation

pip install mcp pandas

Usage

Run the server

python server.py

Claude Desktop config

{
  "mcpServers": {
    "csv-analytics": {
      "command": "python",
      "args": ["/path/to/csv-analytics-mcp/server.py"]
    }
  }
}

Workflow

1. Load a CSV file (stored in memory by name)
2. Explore with describe_columns
3. Query with filters, sorting, column selection
4. Aggregate or pivot for summaries
5. Export chart-ready data for visualization

Example calls

Load a CSV file:

Tool: load_csv
Input: {"file_path": "/Users/me/data/sales.csv"}
Output: {"name": "sales", "rows": 15432, "columns": ["date", "product", "region", "amount", "quantity"], "dtypes": {"amount": "float64", "quantity": "int64"}}

Describe columns:

Tool: describe_columns
Input: {"name": "sales"}
Output: {"columns": {"amount": {"mean": 245.32, "std": 89.10, "min": 5.00, "max": 2500.00, "median": 210.50}, "region": {"unique": 4, "top_values": {"North": 4210, "South": 3890, ...}}}}

Query with filters:

Tool: query_data
Input: {"name": "sales", "filter_expr": "amount > 500 and region == 'North'", "sort_by": "amount", "ascending": false, "limit": 10}
Output: {"rows": [...], "row_count": 10, "total_rows": 312}

Aggregate:

Tool: aggregate
Input: {"name": "sales", "group_by": ["region"], "metrics": {"amount": "sum", "quantity": "mean"}}
Output: {"rows": [{"region": "North", "amount": 1234567.89, "quantity": 12.3}, ...]}

Pivot table:

Tool: pivot_table
Input: {"name": "sales", "index": "region", "columns": "product", "values": "amount", "aggfunc": "sum"}
Output: {"pivot": {"North": {"Widget A": 50000, "Widget B": 32000}, "South": {"Widget A": 45000, ...}}}

Export for charting:

Tool: export_chart_data
Input: {"name": "sales", "x_column": "region", "y_columns": ["amount"], "chart_type": "bar"}
Output: {"chart_type": "bar", "labels": ["North", "South", "East", "West"], "datasets": [{"label": "amount", "data": [1234567, 987654, ...]}]}

Pricing

License

MIT