Pipetable — Claude Code MCP server for local data
Pipetable is an MCP server for Claude Code that lets an agent query local data files with SQL. It comes into play when a task involves inspecting tables or querying their contents.
MIT ★ 5 updated 2026-10-30
Open on GitHubWhat Pipetable does
Pipetable scans a folder of data files and exposes tools for listing datasets, checking schemas, and running DuckDB SQL. It handles CSV, Parquet, JSON, NDJSON, TSV, and Excel files. The files stay on the machine.
Scanning reaches 3 directory levels, and files can be up to 2GB. Hidden files and common directories such as node_modules, target, and .git are skipped. Natural-language queries in the CLI need Ollama running locally or an Anthropic or OpenAI-compatible API configuration. SQL and MCP work without those model services.
What it helps with
- query local CSV and Excel files with SQL
- inspect a dataset’s columns and sample rows
- group results from several data files
- search records stored in Parquet or JSON
Example prompts for the agent
- Find the regions with the highest revenue in
~/data/. - Show the schema and a few rows from the sales table.
- Group sales by region using a SQL query.
How to install Pipetable
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Pipetable
How do I install Pipetable in Claude Code?
Install the program with cargo install pipetable, then register its MCP server using claude mcp add pipetable pipetable mcp.
What does Pipetable do?
It lets an agent inspect local data schemas and run DuckDB SQL queries against the files.
What does Pipetable need to run?
SQL and MCP do not require a model or API key. Natural-language queries in the CLI need Ollama or an Anthropic or OpenAI-compatible API.