I Turned an S3 Bucket Into a Git Remote—No Server Required

Description and use

git3 turns an S3 bucket into a Git remote without a server, using standard Git commands and AWS credentials.

Problem

Developers need a simple, server‑less Git remote that uses existing AWS infrastructure, avoiding extra services for storage, authentication, and maintenance.

Outcome

git3 provides a server‑less Git remote that works with standard Git commands, uses AWS credentials for authentication, stores immutable packfiles in S3, and supports atomic ref updates via conditional writes. It requires no additional infrastructure and preserves local Git repositories as normal.

Workflow

• Install git3 via one‑line curl script
• Configure AWS credentials (default, profile, or env vars)
• Clone a repository: git clone s3://bucket/repos/example
• Add remote and push first branch: git remote add origin s3://bucket/repos/example && git push -u origin HEAD:refs/heads/main
• Use standard Git commands (fetch, pull, push, delete, force, atomic, submodule)
• Run maintenance commands: git s3 doctor, git s3 fsck, git s3 gc, git s3 set-head

Evidence

git3 supports the everyday full‑repository workflows I wanted it to feel boringly normal to use with.

Normal clone, fetch, push, and maintenance operations do not need s3:ListBucket or s3:DeleteObject.

git3 is a storage and synchronization tool, not a GitHub or GitLab replacement.