Proto-platform
I ship a lot of prototypes — and every time, getting them online took longer than the prototype itself: manifests, CI, secrets, certificates. The proto platform handles that for me: pick a repo, paste the .env, click Deploy. A few minutes later the app runs at its own address.
- For
- Prototypes & small apps
- Runs on
- k3s · Hetzner
- Operated by
- Web UI or Claude (MCP)
- Status
- In trial
From repo to URL in four steps
Pick a repo
From the list of GitHub repos, or by URL. A Dockerfile is enough.
Paste .env
Values are encrypted in the browser before they're sent.
Deploy
Build, rollout, certificate — you watch it happen live.
To prod
If it's any good: the very same image moves to your own domain.
Simple to use, hard to break
Detect, don't configure
Services, ports, database and required secrets are read from Dockerfile, docker-compose and .env.example.
Secrets: paste once
Values are sealed in the browser. After that nobody can display them again — not even the platform.
proto and prod
Every push lands on proto. “To prod” promotes the identical image. Rollback is one click.
Never roll out something broken
If a rollout fails, the old version keeps running. Missing secrets block prod — and say what's missing.
Claude can do it too
Everything the UI can do is also available via MCP. Claude deploys and reads logs, but never sees a secret value.
Add-ons when needed
Postgres, Redis, AI access with a budget per app, Stripe in test mode, nightly backups.
What it looks like
The screenshots show demo data. Click to enlarge.
Roughly how it works
01
You or Claude
Web UI or MCP connector.
02
Gateway
Knows projects, releases and the sealed secrets.
03
Build
GitHub Actions builds the images on every push.
04
Cluster
k3s at Hetzner, one namespace each for proto and prod.
05
Online
Traefik + Let's Encrypt: proto address and own domain.
The platform is new: developed in a few days in October 2026, AI-first with Claude Code. The first apps are running on it; my existing projects are moving over next.