SQL CSV Chomper

SQL tools for you. SQL tools for your LLM.

Stop staring at raw CSV text. Stop copy-pasting into Excel. Stop writing throwaway Python scripts just to peek at your data. SQL CSV Chomper turns VS Code into a fast, lightweight SQL workbench for your data files — powered by DuckDB WASM running entirely in your browser. No server, no database, no nonsense.

Open any CSV. Write SQL. Get results. Chomp chomp.

Why?

Working with CSV files in VS Code has always been painful. Need to filter 500K rows? Join two CSVs? Edit a single value and save? Your options were Excel, a Python REPL, or a full database setup. None of those live inside your editor.

And if you're working with an AI assistant? Every CSV operation means another round-trip — "write me a script to filter this", "now save it back", "actually change that column name first". It's slow.

Chomper fixes both problems. You get a visual SQL IDE with filtering, editing, and sorting. Your AI gets an MCP server with direct SQL access to your files. Everyone chomps faster.

Features

Query 500K+ rows without breaking a sweat

Data loads in 10,000-row chunks with automatic pagination as you scroll. The status bar shows exactly where you are — no freezing, no waiting.

Excel-style column filtering

Click the filter icon on any column header to get a rich filter panel:

• Value checkboxes with frequency bars and counts
• Search within the value list (Select All respects your search)
• Null statistics and distinct count
• Active filters appear as chips above the table — click to remove

Date & time intelligence

Date columns get special treatment — histogram buckets by hour, day, week, month, or year. Click any bucket to instantly filter to that time range.

Edit cells inline and save back

Double-click any cell to edit in place. Changes update the in-memory database instantly. Hit Save to write back to disk, or Save As to export with custom delimiters, quoting, and formatting.

SQL intellisense

Full autocomplete for table names, column names (auto-quoted when they contain spaces), SQL keywords, and 60+ DuckDB built-in functions.

Scroll through massive files

Virtual scrolling keeps things smooth even with hundreds of thousands of rows. Load more data automatically as you scroll, or run a query to narrow things down.

Resizable columns

Drag any column edge to widen it — handy when one field holds a long description and the rest are short codes. Double-click a resize handle to snap that column back to a best fit.

Column operations

Right-click any column header to rename, insert, or delete columns. Load multiple files and JOIN across them — the schema explorer shows every table with columns, types, unique counts, null percentages, and the longest value in each column. Hover a column for the full breakdown: distinct, null, whitespace-only and numeric-looking value counts.

Find & Replace

Search and replace across any column with case-sensitive and regex support. Live match count shows how many values will change before you commit.

Your file's format survives the round trip

Chomper detects the delimiter, quoting, escape character and header row when it opens a file, and reuses them when you save. A pipe-delimited or tab-delimited file goes back to disk the way it arrived — it is never quietly rewritten as comma-delimited. Fields that contain the delimiter are re-quoted on the way out, so they still read back correctly.

Load Options — when detection gets it wrong

Some files defeat auto-detection. Click Load Options... to re-read the file with exactly what you specify:

• Column delimiter: comma, tab, pipe, semicolon, or custom
• Row delimiter: LF, CRLF, or CR
• Header: tell it whether the first row is a header
• Quoting: set or clear the quote and escape characters
• Encoding: UTF-8, Latin-1, or UTF-16
• Also: skip leading rows, read every column as text, or skip malformed rows

Save As with options

Export your data exactly how you need it — defaults follow the source file's own format:

• Delimiter: comma, tab, pipe, semicolon, or custom
• Quoting: always, as needed, or never
• Options: include/exclude header, add row numbers
• Format: .csv, .tsv, .txt

MCP Server — Let your AI chomp too

SQL CSV Chomper includes a bundled MCP (Model Context Protocol) server so AI assistants can work with your CSV files directly. No more "write me a Python script to filter column X" — just let them query.

The MCP server is also available as a standalone npm package: sql-csv-chomper-mcp

GitHub Copilot

The MCP server registers automatically in VS Code 1.99+. Copilot can load files, run queries, edit data, and push results into the visual editor.

Claude Code

The MCP server is published as an npm package — no path wrangling needed. Works on macOS, Linux, and Windows.

Option 1 — From VS Code (easiest):

1. Cmd+Shift+P (or Ctrl+Shift+P)
2. Search "SQL CSV Chomper: Configure MCP for Claude CLI"
3. Choose User (all projects) or Project (this workspace)
4. Done — Claude can now chomp your CSVs

Option 2 — Terminal (any OS):

# All projects
claude mcp add sql-csv-chomper --scope user -- npx -y sql-csv-chomper-mcp

# This project only
claude mcp add sql-csv-chomper --scope project -- npx -y sql-csv-chomper-mcp

Upgrading? Remove the old MCP first:

claude mcp remove sql-csv-chomper

MCP tools available

Empty fields read as NULL. DuckDB treats an empty CSV field — and a quoted empty field — as NULL rather than an empty string, so query them with IS NULL (= '' won't match). profile_table separates true nulls from whitespace-only values, which DuckDB does keep distinct.

Keyboard shortcuts

Supported files

The delimiter is detected from the file's contents, not its extension, so an oddly-named file still opens correctly — and Load Options... is there for the ones that defeat detection. Other delimited files (.dat and friends) can be pulled in with + Open File from the schema explorer.

How it works

• DuckDB WASM runs entirely in your browser — zero external dependencies
• Files load into an in-memory DuckDB instance when opened
• SQL executes at native speed against the in-memory database
• Edits modify the in-memory table; Save writes back to disk
• The MCP server runs a separate DuckDB instance for AI workflows

Requirements

• VS Code 1.85+
• Node.js (for the MCP server only)

License

MIT