Query CSV files without uploading to the cloud

ChatGPT truncates large CSVs. Uploading sensitive data to a cloud service is risky. Pipetable runs on your machine — DuckDB reads your files in place, with no size limit and no upload.

Install

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

Scan your folder

pipetable ./data/

Pipetable scans all CSV, Parquet, JSON, Excel and TSV files up to 3 folders deep. Each file becomes a queryable DuckDB view named after the file.

Scanning a folder, running SQL, and querying in plain English

Run SQL queries

SELECT region, SUM(revenue) AS total
FROM sales
GROUP BY 1
ORDER BY 2 DESC

DuckDB handles files that would crash Excel or pandas — 1GB CSVs run in milliseconds.

Ask in plain English

Set your preferred LLM provider and type questions naturally:

# Claude (best quality)
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export OPENAI_API_KEY=sk-...

# Or run Ollama locally (no API key needed)
ollama serve && ollama pull qwen2.5-coder:1.5b

> show me top 5 customers by revenue
> which region had the highest growth last quarter?
> how many orders were above £10,000?

Pipe results for follow-up queries

Every query saves its result as _last. Query the filtered set without re-reading the original file:

> SELECT * FROM sales WHERE region = 'EU'
→ piped as _last

> show me top 3 reps from _last

Why not just upload to ChatGPT?

• ChatGPT truncates files over ~50,000 rows
• Your data leaves your machine and lives on OpenAI's servers
• Results are generated — Pipetable returns real query output
• Pipetable handles files up to 2GB with no degradation

More guides