Download

One binary. No account.

Install it, point it at your data, and start querying. The Community build is free and needs nothing else — no Docker, no JVM, no sign-up.

Version v0.30.0
Released 23 Sep 2026
License Community — free

Release notes and every previous build live on GitHub.

Detected: macOS Bufflehead 0.30.0 for macOS Bufflehead-0.30.0-macOS.dmg
Apple Silicon & Intel · 120.9 MB
Download the .dmg Apple Silicon & Intel. Every build is listed below.

All platforms.

v0.30.0 · 4 assets · release notes on GitHub
New in v0.30.0

Private databases. Whole folders. Easier updates.

SSH tunnels, built in

Reach Postgres or MySQL through your SSH jump host without keeping a separate terminal tunnel open. Use your SSH agent, a key file or a password. Host keys are checked against known_hosts.

Drop a folder and explore

Drag a folder into Bufflehead to open a connection. Browse file patterns such as *.parquet and *.csv with file counts and columns, then click a pattern to query its matching files together.

Check for Updates

See release notes and download updates inside the app, with file-size and SHA-256 verification. Supported macOS installations offer Install and Restart; Windows and Linux hand you the verified package to install.

Also in this release: a File / Help menu bar on Windows, a database switcher that stays anchored to its breadcrumb, and improved Windows folder paths. Read the full release notes.

The app checks GitHub for updates on startup. To disable automatic checks, launch with BUFFLEHEAD_NO_UPDATE_CHECK=1. Older versions can upgrade using the downloads above.

Once it's installed.

01
Add a connection Drop a folder of Parquet or CSV, connect S3, or add Postgres or MySQL. Use an SSH tunnel for private Postgres or MySQL, or AWS SSM for Postgres.
02
Explore it in the workbench Schema inspector, SQL editor, data grid and Parquet metadata — all in the app.
03
Point your AI tools at it The gateway listens on localhost. Your AI tools query the same connections without a Bufflehead cloud in between. Data shared with an AI provider follows that provider's policy.

Or build it yourself.

The core is open source. Clone it, read it, compile it — the binary you run is the code you audited.

$ go install graphics.gd/cmd/gd@release
$ git clone github.com/kyleparisi/bufflehead
$ gd run
View on GitHub Enterprise features need a commercial license.