By ByteBot
A new open-source platform called Cloud in a Bottle launched this week with a thesis the self-hosting world has been circling around for a decade without acting on: the hardest part of self-hosting isn’t installing software — it’s living with it afterwards. Built by Imbue and released under AGPL-3.0 on September 5, it puts unified single sign-on, sandboxed containers, and a single management dashboard at the center. The result looks less like a sysadmin tool and more like an app store for a server you actually own.
The Problem Nobody Solved
Every self-hosting platform since 2012 has solved the install problem. What none of them solved, however, is the after. You get Nextcloud running, then discover its calendar sync conflicts with your Jellyfin tokens. Your VaultWarden instance uses a different login than your Forgejo. You update one container and something else breaks. In short, the implicit promise of self-hosting — “you own your stack” — buries the operational reality: you are now the IT department for your own life.
Cloud in a Bottle’s central bet is on unified authentication. Log into your instance once and you are logged into every app — Nextcloud, Jellyfin, VaultWarden, Open WebUI, SearXNG, Forgejo. Moreover, apps can share data via defined service interfaces. This is the feature that changes the daily experience. Not the installer.
How Self-Hosting Gets Its App Store
At its core, Cloud in a Bottle is an Ubuntu machine with a Python control plane that routes HTTP requests to rootless Podman containers — one per app. Caddy handles HTTPS. Additionally, CoreDNS provides wildcard subdomains so every app gets its own clean URL. Apps are defined via cloudinabottle.toml manifests, meaning any container-ready application can integrate without forking or modification.
Hardware options are deliberately flexible: run it on an old laptop, bring your own VPS, or pay Imbue a month for a managed instance. The launch post outlines the full architecture. Crucially, the AGPL-3.0 license keeps the source open while the managed hosting option gives Imbue real revenue — something the self-hosting ecosystem has long lacked.
Why Sandstorm Failed — and How This Tries to Be Different
Inevitably, the comparison to Coolify, Cloudron, and Yunohost follows. Coolify gives you one-click app deployment but each app maintains its own login. Cloudron does unified auth properly but starts at 2 per month (license plus VPS). Yunohost skips containerization entirely, meaning one compromised app can reach all your data. And Sandstorm, which attempted exactly what Cloud in a Bottle is now attempting, is effectively dead.
Sandstorm’s failure is instructive. It required apps to be specifically ported to its security model, which created a chicken-and-egg problem: few apps supported it, so few users adopted it, and consequently developers did not prioritize it. Cloud in a Bottle sidesteps this by using standard containers — no fork required. However, the real test is ecosystem momentum. The platform launched with 202 GitHub stars. Coolify has 25,000. The GitHub repository shows active development, but the app catalog depends entirely on community contributions to grow.
The Astroturfing Problem
Before the launch, Imbue team members posted at least eleven GitHub issues to unrelated repositories requesting cloudinabottle.toml support — without disclosing they were Imbue employees. The community noticed immediately. The Hacker News discussion (429 points, 205 comments) called it out clearly, and the creator acknowledged it was a mistake.
Still, this is a real problem, not merely a PR misstep. Self-hosting platforms run on community trust more than any other category of open-source project. Specifically, you are asking developers to put their personal data, their family’s photos, their password manager on your software. Starting with a trust deficit — even a small one, even one you own up to — is a harder hole to climb out of than it appears. The community memory is long.
The Bottom Line
Cloud in a Bottle is pursuing the right idea. The self-hosting ecosystem has needed unified auth and cross-app integration for years. Furthermore, the managed hosting model at per month is the correct approach to avoid Sandstorm’s fate, and the technical architecture is clean enough that app developers face no additional burden. Whether it ultimately succeeds, however, depends entirely on whether the community builds the ecosystem — and that is a trust and momentum problem, not an architectural one. Imbue handed itself a harder start than it needed to.
