Branch the whole database, not a dump of it
fox branch create qa clones every table, index and row in seconds, whatever the size. The branch is fully isolated: its writes never touch main, and main’s never touch it.
You pay only for what you change
A branch is a copy-on-write clone of main served by its own Postgres: ZFS on macOS and Linux, btrfs on Windows. Blocks are shared with the parent until the branch writes to them.
So a copy of a large database costs almost nothing at first, and grows only by what it changes.
- Idle branches suspend themselves and wake on the next connection.
- Connect to …:6432/qa: the database name is the branch.
- Branches belong to whoever made them; others get “not found”.
Create, reset, compare, sleep
Create
A complete copy of main in seconds, with its own connection string.
Reset
Throw away a branch’s changes and start again from its parent.
Diff
See the schema changes that separate two branches, straight from the Blackbox.
Suspend
Stop a branch and keep its data; the next connection wakes it.
The whole workflow in six commands
Every command also exists in the web console and the REST API.
$ fox branch create qa $ fox connect qa $ fox branch diff main qa $ fox branch reset qa $ fox branch suspend qa $ fox branch delete qa
What teams branch for
A database per pull request
Branch main for each PR, point the preview at it, delete it on merge.
Migration rehearsal
Run the risky migration on a full-size copy first, and read what it changed.
A sandbox per agent
Each AI agent gets its own copy and a key that opens only that copy.
Make your first copy in seconds.
fox branch create qa, on the database you already have.
$ curl -fsSL https://get.thefoxbyte.com/install.sh | sh