The DuckDB app and MCP server
Open, convert, combine and analyse big CSV, Excel, Parquet and JSON files — millions of rows in seconds, on your own computer.
Free · no account · 3 free uses of every tool · also for Mac (Intel), Windows, Linux
See the columns, a preview of the first rows, and per-column stats (types, empty cells, min/max, unique values) of any CSV, Excel, Parquet or JSON file — even huge ones.
Turn CSV into Excel, Excel into CSV, or either into Parquet or JSON.
Stack many CSV files (e.g. monthly exports) into one table. Columns are matched by name, so files with different column orders still line up.
Save a copy with exact duplicate rows removed.
Answer questions about your data with SQL — totals, filters, joins across files. Refer to files by path in single quotes, e.g. SELECT region, sum(amount) FROM '/path/sales.csv' GROUP BY region.
GUI Hub gives your assistant these 5 tools. Install it once (below) and ask in plain words.
| MCP tool | What it does | Inputs |
|---|---|---|
| duckdb_summarize | Inspect a local data file (CSV/TSV, Excel .xlsx, Parquet, JSON) without loading it into the chat: returns the first rows and per-column statistics (type, nulls, min, max, distinct count) as Markdown tables. Use this before writing a query. | input* string rows number |
| duckdb_convert | Convert a local data file between CSV, Excel (.xlsx), Parquet and JSON. Writes a new file; the original is not changed. | input* string format string out* string |
| duckdb_combine | Combine several local CSV files into one file (rows stacked, columns matched by name; missing columns become empty). Output format CSV, Excel, Parquet or JSON. | inputs* array format string out* string |
| duckdb_dedupe | Write a copy of a local data file with exact duplicate rows removed (CSV, Excel, Parquet or JSON in; chosen format out). | input* string format string out* string |
| duckdb_query | Run DuckDB SQL over local data files and get the answer as a Markdown table. Read files by quoted path: SELECT * FROM '/abs/path/file.csv' (also .xlsx, .parquet, .json; globs like '/dir/*.csv'). Use for totals, filters, group-bys, joins across files on data too big to paste into chat. Results are capped by the row limit; to save results use COPY (query) TO '/abs/out.xlsx' (FORMAT xlsx, HEADER true). | sql* string maxRows string |
* required
Free to try — 3 full uses of every tool.
Download the one-file extension, double-click it, press Install. Done — ask Claude for a file job.
Download gui-hub.mcpbAdd an MCP server with this command (needs Node.js 18+):
npx -y github:MuttyRab/MuttyRab.github.io
Claude Code:
claude mcp add gui-hub -- npx -y github:MuttyRab/MuttyRab.github.io
Point-and-click forms for every tool, plus a Use with AI button that connects Claude Desktop or Cursor for you.
Download for WindowsNew independent apps get this one-time caution. The full source is public on GitHub.
“Look inside a file” runs the official DuckDB 2.0.0 program with arguments passed directly — never through a shell.
duckdb -markdown -c "SELECT * FROM '/Users/you/Movies/in.mov' LIMIT 10; SUMMARIZE SELECT * FROM '/Users/you/Movies/in.mov';"
Use Spreadsheet & Data Toolkit 3 times free. $9 once unlocks all of GUI Hub Data.
Every tool works 3 times free first. Secure checkout by Polar; your license key arrives by email — paste it into the app, or just give it to your AI. 30-day refunds.
Powered by DuckDB 2.0.0 (MIT), downloaded from its official release and verified by SHA-256. GUI Hub is independent and not affiliated with DuckDB.