In development. Not yet available.
Whether it's the media server in your closet or the SaaS that pays your salary, grund takes any app from a bare machine to deployed, backed up and recoverable. Start on one box. Grow to a fleet without starting over.
A container image or a Dockerfile: an open-source tool you self-host, or the product you are building.
A mini PC, an old desktop, a rack in the basement or a rented dedicated server.
Add machines when it matters. The apps you deployed on day one keep running, now highly available.
Your hardware, your data, your git. A standard cluster that keeps running without us.
Every level builds on the last. Stop wherever you are happy: a homelab can live at level three for years. A SaaS keeps climbing.
No cloud account, no platform team. One box with an internet connection is enough to begin.
grund installs a pinned, tested stack: a cluster, ingress with automatic TLS, metrics, and backups to storage you choose.
$ grund init --server nuc.home.lan
✓ cluster ready on nuc
✓ ingress and certificates
✓ metrics and alerts
✓ backups to s3://home-backups
Point grund at an image or a Dockerfile. It gets a hostname, a certificate and health checks, and restarts when it falls over.
$ grund deploy ghcr.io/immich-app/immich-server \
--domain photos.home.example
✓ healthy in 24 s
✓ https://photos.home.example
Databases stream their changes off-site. Every night grund restores them somewhere safe and checks the result, so a dead disk is an afternoon, not a disaster.
Join two more servers and grund spreads your apps and replicas across them. Nothing is redeployed from scratch; one machine can now fail without taking you down.
$ grund add server srv-2 srv-3
✓ 3 servers in quorum
✓ replicas rebalanced
✓ can lose any 1 server
Releases are promoted by digest through health gates. A bad one rolls back on its own before your customers notice, and every guarantee comes with evidence.
Commands and output are illustrative. grund is in development and its interface may change.
The same path serves a family photo library and a product with paying customers. Two examples of where the journey leads.
One mini PC, level 3.
Three dedicated servers, level 5.
Example setups. grund is in development.
A backup nobody has restored is a hope. grund tests its own promises on a schedule and shows you what it measured, at every level of the journey.
grund automates a stack you could run yourself. Nothing about it depends on us staying around, or on your hardware staying where it is.
$ kubectl get nodes
NAME STATUS ROLES
srv-1 Ready control-plane
srv-2 Ready control-plane
srv-3 Ready control-plane
k3s, Flux and CloudNativePG. No proprietary runtime you cannot inspect.
Plain manifests and encrypted secrets in your repository, applied by Flux.
New servers, no rewrite
From the closet to a rented rack, or back. Point grund at new machines and restore onto them.
No SSH keys, no remote commands
An agent on your machines pulls signed intents and verifies them. We never hold your keys.
Made for
Not what it promises
The cost, up front
Start on hardware you already own, for nothing but electricity.
Real high availability needs three machines, a quorum. Rented from a provider like Hetzner that is roughly €290 a month before storage, often a fraction of the cloud bill it replaces.
Follow along on GitHub, where the code appears as it lands.