Your own private cloud — passwords, photos, files, and calendars on hardware you control.
Website · Get started · Documentation · Changelog
My Own Suite (MOS) turns one machine — a mini-PC at home or a small rented cloud server — into a private replacement for the everyday cloud: install it once, then add the open-source apps you want from a curated catalog in a couple of clicks. MOS handles the databases, networking, HTTPS, and secure wiring in the background; you get one owner UI and apps that just work, with your data staying on hardware you control.
- Privacy you can check, not marketing. Every catalog app is assessed before it enters the catalog and carries an A-to-D privacy posture grade — telemetry, external services, accounts, data processing, and policies, each with published evidence. How assessments work.
- A signed catalog and pinned packages. Apps build from digest-pinned recipes; your server verifies what it installs, and updates show you what changes — including privacy changes — before you apply them.
- One-button backups with verified restore. Whole-suite backups to storage you control, checked for integrity before they're ever restored.
- Apps that plug into each other. Connect Seafile to ONLYOFFICE and documents open for editing in the browser — MOS exchanges the secrets and wires the network.
- Not a walled garden. Paste any GitHub repository URL and MOS previews what the package asks for before anything installs. External apps stay labelled unverified and run restricted — and publishers can make their own apps MOS-installable with the public packaging workflow.
On a fresh Ubuntu 24.04 server (your provider's browser console or SSH):
curl -fsSL https://get.myownsuite.org | sudo bashOr flash the USB installer for a dedicated machine at home. The getting-started guide walks through both paths.
Status: early beta. The platform is under active development and tested end-to-end, but expect rough edges and breaking releases while pre-1.0.
This repository is the whole platform — control plane, app packages, installers, docs site, and the review workflows behind the privacy assessments.
| Path | Purpose |
|---|---|
suite-manager/ |
MOS web UI, backend API, owner setup, app catalog, settings, backups, and updates. |
apps/ |
Self-contained MOS app packages with manifests, Dockerfiles, icons, setup guides, and technical notes. |
system-agents/ |
Narrow host-owned agents for app runtime, Homepage, HTTPS, backups, updates, and lab reset. |
infrastructure/ |
Shared runtime substrate for Caddy, Homepage defaults, installer support, and control-plane contracts. |
scripts/ |
MOS installer renderers, smoke harnesses, reconciliation, branding sync, and developer commands. |
skills/ |
Public, versioned workflows for adding apps, updating apps, and assessing app privacy. |
shared/ |
Cross-process contracts used by Suite Manager and host agents. |
test/ |
Deterministic unit tests and browser/E2E harnesses. |
site/ |
MOS public landing page and end-user documentation source; deployed to Cloudflare Pages from main and staging via GitHub Actions. |
site-mos1-reference/ |
Isolated previous-site source retained only as frozen rollback/reference material; not built or deployed. |
Install dependencies, build the Suite Manager frontend, and run the local control plane:
npm install
npm run hooks:install
npm run devnpm run hooks:install wires the local git hooks in .githooks/ that block direct commits/pushes on main.
Open http://home.localhost:3100/suite-manager/.
Useful deterministic commands:
npm test
npm run typecheck
npm run build:client
npm run install:render -- --target json
npm run release:check
npm run build # currently configured deployment build; see the cutover checklistBrowser and infrastructure smoke commands are intentionally human-run because they are noisy or can create paid/destructive resources. See scripts/README.md and test/README.md.
| Need | Go here |
|---|---|
| Documentation ownership | docs/README.md |
| How to contribute | CONTRIBUTING.md |
| Reporting a vulnerability | SECURITY.md |
| Architecture decisions | docs/decisions.md |
| Release process | RELEASING.md |
| Release notes | CHANGELOG.md |
| Suite Manager technical notes | suite-manager/README.md |
| App package technical notes | apps/<app>/README.md |
| App review workflows | skills/README.md |
| Host-agent notes | system-agents/README.md |
| Operator/developer scripts | scripts/README.md |
| Test harness notes | test/README.md |
For day-to-day prototyping, use staging as the integration branch and reserve main for stable release-ready batches. Agent and contributor workflow rules live in AGENTS.md.
My Own Suite is free and open-source software licensed under the GNU Affero General Public License v3.0 only. You may use, modify, distribute, and commercially host it under the license terms; modified versions offered over a network must offer their corresponding source code to their users.
The My Own Suite, MOS, and Funkyton names and logos are not granted for use as the identity of a modified distribution or competing service. Third-party application names, logos, screenshots, and other assets remain the property of their respective owners.
This project uses AI-assisted development tools, including Codex and Claude Code, to support rapid prototyping during the beta phase. Much of the code was written with that assistance, and we would rather explain how it is verified than leave you guessing.
Humans decide what gets built. Requirements, architecture, technical constraints, security boundaries, and acceptance criteria are human decisions. AI agents are assigned small, clearly scoped tasks. They are not handed broad feature requirements and left to design, implement, and approve a complete solution unsupervised.
Nothing ships that a human has not driven by hand. No feature reaches a release until it has been exercised on a real deployment — not a mocked test, an actual install. That is why so much of this repository is verification tooling rather than product: a Hyper-V harness that builds the USB installer and installs the platform from scratch on a virtual machine, an automated DigitalOcean harness that does the same on a real cloud server, a browser E2E suite that walks the genuine install-to-apps flow with no test-only bypasses, and a deterministic unit suite that runs on every push and pull request. The point of all of it is to make hands-on human validation cheap enough to do constantly instead of occasionally, and it is done constantly.
AI is not treated as an authority on whether the software works. It is used for implementation, troubleshooting, research, and exploring options. Whether MOS behaves correctly is decided by a human running it.
The honest limit. This process validates behaviour — that the platform installs, upgrades, backs up, restores, and fails safely on real machines. It is not the same as a line-by-line human audit of every module, and we do not claim it is. That full audit is what the alpha is for: if the project proves its worth, every module goes through human verification before we use that word. The gaps we already know about are recorded in the Alpha gate section of docs/roadmap.md and in SECURITY.md, rather than left for someone to discover.
The maintainers remain responsible for every change in the project, regardless of whether AI contributed to its implementation.