Serverless Git for repositories that outgrew ordinary Git.

Beyondnote Technology Inc builds Crab, a serverless Git remote helper and large-file workflow for teams that want to use their own cloud storage instead of running a separate Git LFS service.

The project is centered on a Rust CLI that can act as git-remote-crab, a Git filter driver, and a set of repository maintenance commands. Crab also includes docs, a desktop app, and optional enterprise services for authentication, caching, and operational support.

Large files make ordinary Git slow and expensive. Git LFS helps, but it usually introduces a hosted endpoint or service contract between developers and their data.

Crab takes a different shape: Git stores small pointer blobs, while actual file content is chunked, deduplicated, compressed, and stored in object storage that the user or organization controls. Developers can clone quickly, hydrate only the files they need, and keep standard Git commands in the loop.

We optimize for boring operational boundaries: resumable uploads, content-addressed chunks, byte-identical reconstruction, explicit cloud credentials, local caches, and repository-scoped metadata.

The CLI supports AWS S3, Google Cloud Storage, Azure Blob Storage, and S3-compatible systems. For larger teams, optional Crab Auth and cache services can be deployed by the customer to add identity-based authorization, audit trails, and shared cache acceleration.

Crab is designed so the normal data path does not require Crab to host repository contents. Your configured cloud bucket is the origin for repository data, and your identity provider or cloud IAM policy decides who can access it.

When users contact us, use the website, download releases, or opt in to enterprise support, those interactions have their own privacy and service terms. The product boundary is intentional: repository data should stay where the customer expects it to stay.