Introduction

Quick Installation

Docker

docker run --name duck-ui -p 5522:5522 \
  ghcr.io/caioricciuti/duck-ui:latest

Access at: http://localhost:5522

Full Installation Guide

What is Duck-UI?

Duck-UI is an open source workbench for DuckDB, the fast in-process analytical database. It runs DuckDB as WebAssembly inside your browser tab, so you get a SQL editor, notebooks, charts and dashboards without any server infrastructure. Your data never leaves the tab.

Key Features

No Server Required: DuckDB runs in your browser via WebAssembly
SQL Editor: CodeMirror with completion from your catalog, DuckDB dialect formatting, run the statement under the cursor, and inline error marks where DuckDB points
Results: Virtualized grid with sort, filter, search, cell selection and copy, column statistics, and exports to CSV, JSON, XLSX and Parquet
Notebooks: SQL, Python (Pyodide) and markdown cells, each with its own result and chart
Markdown Dashboards: Reports written as markdown with live SQL, charts, and input variables (Dashboards)
Charts: Bar, grouped bar, stacked bar, line, area, stacked area, pie, donut and scatter, with PNG export (Charts)
Live Sessions: Real-time collaboration between browsers, peer-to-peer, with shared data and visible cursors (Live Sessions)
Duck Brain: Natural language to SQL with a local server (Ollama, LM Studio), OpenAI, Anthropic, or a model running inside the browser (Duck Brain)
Import Anything: Load CSV, JSON, Parquet, Arrow, XLSX and .duckdb files from your computer or from URLs
Folder Access: Mount folders from your computer; the selection persists across browser sessions (Folder Access)
Persistent Storage: Save databases using OPFS for data that persists across sessions
External Connections: Connect to remote DuckDB HTTP servers when needed
History: Every run is kept with duration and row count, searchable across reloads
Share and Embed: Encode an analysis into a URL, or embed a live query in any page (Embedding)
Works on a phone: Below 768px the rail moves to the bottom and the explorer becomes a drawer
Installable: Duck-UI is a PWA and works offline after the first visit
Privacy First: All data processing happens client-side

How the app is organised

Queries, notebooks and dashboards open as tabs in the workspace. Everything else is a page with its own sidebar, reached from the groups on the left rail:

Rail group	Pages
Query	The workspace with your tabs
Library	Dashboards, Saved queries, History
Data	Connections, Extensions
Settings	Profile, General, AI, Performance, Project

Pages live in the URL (?page=settings&section=ai), so a reload or a bookmark lands on the same place. ⌘K (Ctrl+K on Windows and Linux) opens the command menu, which finds pages, tabs, tables, saved queries, dashboards and recent queries.

Requirements

Modern web browser with WebAssembly support:
- Chrome/Edge 88+
- Firefox 79+
- Safari 14+
Docker (for containerized deployment) or Bun/Node.js 20+ (for building from source)
No database server needed for local WASM mode

Use Cases

Data Analysis

Analyze CSV, JSON, or Parquet files directly in your browser without uploading to a server:

SELECT product, SUM(sales) as total_sales
FROM read_csv('sales_data.csv')
GROUP BY product
ORDER BY total_sales DESC;

Quick Prototyping

Test SQL queries and data transformations without setting up infrastructure:

-- Load data from URL
CREATE TABLE products AS
SELECT * FROM read_parquet('https://example.com/products.parquet');

-- Run analytics
SELECT category, AVG(price) as avg_price
FROM products
GROUP BY category;

Learning SQL

Perfect environment for learning SQL with instant feedback and no setup:

Import sample datasets
Write queries with autocomplete
See results immediately
No database configuration needed

Support

License

Duck-UI is open source software licensed under MIT.