Query Excel files with AI — no upload required

Uploading spreadsheets to ChatGPT or Claude.ai means your data leaves your machine. Pipetable reads Excel files directly from disk — no upload, no size limit, real SQL results.

Install

curl -fsSL https://pipetable.com/install | sh

Point it at your Excel files

pipetable ./reports/

Pipetable automatically detects and loads .xlsx, .xls, and .xlsm files alongside CSV, Parquet and JSON. No conversion needed.

Query with SQL

-- Your Excel sheet becomes a DuckDB table instantly
SELECT product, SUM(revenue) AS total
FROM quarterly_report
GROUP BY 1
ORDER BY 2 DESC
LIMIT 10

Query with natural language

Set ANTHROPIC_API_KEY or OPENAI_API_KEY and ask questions directly:

> show me total sales by region from quarterly_report
> which products had declining revenue quarter over quarter?
> what's the average deal size for enterprise customers?

Use as an MCP server with Claude Code or Cursor

Claude Code can query your Excel files directly once Pipetable is connected:

claude mcp add pipetable pipetable mcp

Then ask Claude: "scan my ~/reports folder and tell me which Excel files are there and what columns they have."

How Excel loading works

Pipetable uses calamine — a pure Rust Excel reader — to load spreadsheets into DuckDB tables. Numeric columns are typed as DOUBLE, everything else as VARCHAR. The first row is treated as headers. Only the first sheet is loaded (multi-sheet support coming).

More guides