Why an MCP server for this
Business reporting
Generate a first-cut analysis report from operational or sales spreadsheets in one call, then refine the narrative in the agent.
Data quality checks
Surface duplicate rows, columns missing over 20% of values and numeric outliers before a model or dashboard touches the data.
Finance and FP&A
Compute totals, distributions and significant correlations on budget and ledger files locally, with no third-party data processing.
AI agents and RAG
Give Claude Desktop, Cursor or any MCP client a callable tool that returns structured statistics and an XLSX instead of hallucinated figures.
What it does
- analyze_table — turn a CSV/XLSX into summary stats, duplicate/missing diagnostics, top-5 category frequencies and pairwise correlations (Pearson |r| ≥ 0.5).
- Numeric columns report min, max, mean, median, sum, standard deviation and missing rate; categorical columns report unique count and top frequencies.
- Optional Excel workbook with sheets for overview, numeric stats, detail, per-column distribution charts (native bar charts) and significant correlations.
- Chinese Markdown insights writeup included as a starting point — explicitly framed as analysis input, not conclusion, so business judgment stays with the user.
- Runs on the Python standard library plus openpyxl — no pandas, so it installs lightly and bundles into small environments.
- health — service status check. 100% local, no network, optional local-only usage counter.
Install and connect
Published to the official MCP Registry as io.github.diyaluo/data-report-mcp and as an OCI image on GHCR. Run locally over stdio:
pip install -e .
data-report-mcpClient configuration (Claude Desktop / Cursor / any MCP client):
{
"mcpServers": {
"data-report": {
"command": "data-report-mcp",
"env": {}
}
}
}Frequently asked questions
What does this MCP server do?
It exposes spreadsheet analysis as an MCP tool, so an MCP-compatible AI agent can take a CSV or XLSX file and return a structured report with summary statistics, data-quality diagnostics, category frequencies, correlations and an optional Excel workbook containing native charts.
Does it upload my spreadsheets?
No. The server runs locally over stdio and makes no network calls. Business and financial spreadsheets stay on the machine, which is why teams handling confidential data use it instead of online analysis tools.
Do I need to install pandas?
No. It is built on the Python standard library plus openpyxl for Excel output, deliberately avoiding a heavy pandas dependency so it installs quickly and works in constrained environments.
Which MCP clients are supported?
Any client supporting the Model Context Protocol over stdio, including Claude Desktop and Cursor. It is published to the official MCP Registry as io.github.diyaluo/data-report-mcp and distributed as an OCI image on GHCR.
Is the generated report a final conclusion?
No. The Markdown insights and statistics are an analysis starting point. Correlation is not causation; the output is designed to feed human business judgment rather than replace it.