A local DuckDB UI — editor, charts, and profiling in one window

Home · Quick start · Editor · charter.yaml · Sources · Agent · Guides · Evals · Audit · Policies · CLI · MCP · Workspace · Desktop · About · FAQ

If you love DuckDB and want a proper UI on top of it — an editor, results you can see, charts, profiling — DataCharter gives you exactly that in one pip install, and keeps everything on your machine.

uvx datacharter serve
# -> a full DuckDB workspace on http://127.0.0.1:8321, demo data included

What you get on top of DuckDB

DuckDB is the engine; DataCharter is the cockpit:

A real SQL editor (Monaco, the VS Code editor) with DuckDB syntax, ⌘/Ctrl+Enter to run, and live results that preview as you type.
One-click charts — auto-detected from your result shape, powered by Vega-Lite, no chart-builder ceremony.
A profiling panel — missing values, distributions, percentiles, outliers, and per-column top-value bars for any result.
Query history and saved queries — every run is kept locally; reopen or share the .sql file from your workspace.
Drag-and-drop files — drop a CSV, Parquet, JSON, or Excel file on the window and it’s queryable instantly.
EXPLAIN and scan estimates — see the plan, and get warned before a query scans a mountain of data.
Exports — CSV, Parquet, JSON, XLSX, one click.

Not just local files: federation

Because it’s DuckDB underneath, one SQL statement can join across sources: local files next to Postgres, MySQL, SQLite, SQL Server, Snowflake, BigQuery, Iceberg, and Delta — configured in a small charter.yaml file with credentials kept in your OS keychain or .env. See Sources for the full list.

And it’s read-only by construction

The engine refuses writes, DDL, and filesystem access at the query layer — so pointing it (or a teammate, or an AI agent) at your production replica can’t do damage. If you later want an agent in the loop, the same workspace adds governed, PII-masked agent access and a governed MCP server — but the UI is a great DuckDB companion even if you never touch the agent side.

Try it

Nothing to configure, nothing global:

uvx datacharter serve      # ephemeral demo workspace

or start a real workspace over your own data in a couple of minutes. It’s Apache-2.0, local-first, and open source on GitHub — come say hi if it makes your DuckDB day nicer.