Serverless Git for large files.
Crab is a serverless Git remote storage solution for teams working with models, datasets, game assets, media, and other large files. Repositories live directly in cloud object storage such as S3, Google Cloud Storage, or Azure Blob Storage, with no servers, databases, or LFS endpoints to operate.
Website · Docs · CLI docs · Desktop docs
- Bring your own bucket — store repository data in cloud storage you control.
- Keep standard Git workflows — clone, commit, push, branch, and review with familiar Git commands.
- Upload less data — content-defined chunking and deduplication send only the pieces that changed.
- Hydrate on demand — clone quickly with lightweight pointer files, then materialize the large files you actually need.
- Use a desktop app when you want one — browse repositories, preview files, hydrate or dehydrate content, diff changes, and work over SSH.
curl -fsSL https://crab.build/install.sh | bash
crab install --global
cd my-project
crab init --storage-provider s3 crab://my-bucket/my-repo
crab setup
crab ship -m "initial commit"Crab supports S3, Google Cloud Storage, and Azure Blob Storage. Configure the matching cloud credentials first, then point Crab at the bucket or container where the repository should live.
- Crab CLI — Rust command-line tool and
git-remote-crabhelper. - Crab Desktop — Electron app with a Rust sidecar for repository browsing, previews, hydration controls, terminal workflows, and remote workspaces.
- Crab SDK — Rust read-side library for repository, snapshot, prefetch, and walk operations.
- Python bindings — PyO3/maturin bindings for Python consumers.
- Docs and guides — product docs, architecture notes, and workflow guides at crab.build/docs.