git-lazy-mount

git lazy-mount repositoryurl mountpointgit-lazy-mount repositoryurl mountpoint

git-lazy-mount mounts a Git repository as a local working tree without first cloning it. Instead of downloading every object up front, it exposes the repository through a FUSE file system and fetches file contents on demand, the moment a file is opened, read, or edited. Files that are never touched are never downloaded, which makes very large repositories usable in seconds and with a fraction of the disk space.Because the mount looks like an ordinary checkout, normal Git commands work inside it: you can edit files, commit, create branches, and push. The project also ships a companion `sgrep` tool that queries a remote code-search index and overlays your uncommitted edits, so searching a huge repository does not require materializing it.It installs as a Git subcommand, so `git lazy-mount` and the underlying `git-lazy-mount` binary are equivalent. The tool is built in Rust and targets Linux, where it relies on the kernel FUSE interface.

git-lazy-mount is Linux only; macOS and Windows are studied but not supported. It requires libfuse3 and a system Git of at least version 2.36. Because file contents are fetched lazily over the network, accessing many files for the first time, or working offline, can be slow or fail, and the mount depends on the remote staying reachable. It is best suited to read-mostly work and targeted edits on large repositories rather than operations that scan the whole tree.

git-lazy-mount was created by Mohsen Azimi (mohsen1) and is dual-licensed under MIT and Apache 2.0. In the author's benchmarks across twenty repositories it used about 1.3 GB versus 23 GB for full clones, roughly eighteen times less, while preserving complete history.

git(1), sshfs(1), mount(8), fusermount(1)

Source code