I built Graft after discovering that deployment tools like Dokploy consume more RAM than my entire t2.micro instance (1GB). The Problem Running Dokploy on AWS t2.micro instances caused constant crashes – the tool itself used 1GB+ RAM before I could even deploy my apps. Manual SSH deployment was tedious and error-prone. The Solution Instead of running an agent on the server, Graft runs locally and SSH's in to handle everything:

Sets up Traefik reverse proxy Pushes source and builds on the server Manages the full deployment lifecycle

Current features (v2.2.3):

graft-hook: Rust-based webhook receiver for CI/CD

1.7MB binary, 15MB RAM during deployment vs. 1GB+ for traditional tools

Automated backups & rollbacks Cloudflare DNS automation Docker Compose passthrough (run any docker-compose command remotely)

Workflow: bash# One-time setup graft init

# Deploy graft sync

# Manage remotely as if local graft logs backend graft restart frontend graft ps Zero-to-deployed in ~15 minutes if your docker-compose already works locally. Installation: Available via Homebrew, Winget, apt, dnf, yay, paru, or shell script. Docs: https://graftdocs.vercel.app GitHub: https://github.com/skssmd/Graft Would love feedback from others running on resource-constrained servers. Open to feature suggestions and contributions.