Quickstart

Keep a file’s version in Git and its bytes in Gat, then restore it in a fresh clone. Before you start, install Gat and open an existing Git repository. Replace assets/model.bin below with a file that Git does not already track. Already using Git LFS or DVC? Follow the migration guide to transfer existing files.

Working with an AI assistant? Connect it to the Gat MCP server so it can search the documentation as you follow this guide.

Initialize Gat

gat init

This prepares the local cache, . Run it once in each clone.

Choose storage

This example uses a local directory with no cloud account or credentials. Choose the commands for your shell:

macOS / Linux
Windows PowerShell

mkdir -p "$HOME/gat-quickstart-storage"
gat remote add origin "file://$HOME/gat-quickstart-storage"
gat remote default origin

$storage = Join-Path $HOME 'gat-quickstart-storage'
New-Item -ItemType Directory -Force $storage | Out-Null
$storageUrl = ([System.Uri]::new($storage)).AbsoluteUri
gat remote add origin $storageUrl
gat remote default origin

A Gat remote stores file content; a Git remote stores commits. Adding a Gat remote never selects a default, even when it is named origin.

Keep this storage directory: the fresh clone will read from it. For shared storage, follow Set up a remote.

Track and commit a file

gat add assets/model.bin
git add gat.lock gat.yaml
git commit -m "Track model with Gat"

The file stays in your working tree. Git commits its path and in gat.lock, plus the storage configuration in gat.yaml.

How does Gat keep the file out of Git?

Gat manages entries in .git/info/exclude so ordinary git add commands skip Gat-managed files. The bytes live in .gat/objects by default. See How Gat works.

Upload the bytes

gat push

When sharing changes with teammates, run gat push before git push so the objects are available before their commits are published.

Restore the file in a fresh clone

From the original repository, clone into a sibling directory:

git clone . ../gat-quickstart-clone
cd ../gat-quickstart-clone
gat init
gat pull

Git supplies the metadata; gat pull downloads the bytes and restores files for the current checkout. This local example works on the same machine.

Open assets/model.bin in the new clone and confirm its contents. gat ls-files lists the tracked paths.

Everyday commands

Task	Command
Record new or edited file content	gat add
List tracked paths	gat ls-files
Compare current tracking metadata with staged metadata	gat status
Compare tracked versions across commits	gat diff
Move a tracked file	gat mv
Stop tracking and delete a file	gat rm
Upload required objects	gat push
Download objects and restore working files	gat pull
Download objects without changing working files	gat fetch
Update working files from the local cache	gat sync

gat status does not inspect file contents. Record edits with gat add; use gat sync --dry-run to preview working-tree reconciliation.

Git hooks run sync after checkout/switch, merge/pull, and rebase/amend. After git reset --hard or git restore, run gat sync yourself. Use gat pull if required objects are missing from the cache.

Next steps

Set up shared storage

Connect cloud storage or a shared directory.

Work with your team

Publish changes and resolve merge conflicts.

Choose a working set

Pull only the paths you need or save a default selection.

Understand automatic sync

Control fetching and working-tree updates after Git operations.

Getting Started

Concepts

Guides

References

Commands

Quickstart

Everyday commands

Next steps

Set up shared storage

Work with your team

Choose a working set

Understand automatic sync

​Everyday commands

​Next steps

Set up shared storage

Work with your team

Choose a working set

Understand automatic sync

Everyday commands

Next steps