Atom Loops¶
Atom Loops is an open-source system for atomic and reproducible deployment of the operating system on embedded and desktop Linux devices. An update is one transaction: download, verify, switch, with automatic rollback on boot failure.
It is not a Linux distribution and not a package manager. It consumes a finished system image and owns only the deploy and rollback lifecycle. It needs no container engine, no registry, and no dedicated partition layout beyond an ESP.
Status
Under active development, not ready for production. The first integration target is Sinty OS, where the code is exercised end to end. Only x86_64 (UEFI) is validated so far. See the roadmap for what is implemented and what is not.
What it guarantees¶
- Atomic. An update is either fully applied or not at all. There is no half-updated state.
- Verifiable. Every image is signed and checked end to end before it is trusted. Trust is the signature, not the transport.
- Reversible. If a new image fails to boot, the boot chain returns to the last good image on its own.
- Reproducible. The system ships as a whole, signed image. What runs is exactly what was built, not the result of changes layered on top.
How it works, in one paragraph¶
A background daemon prepares the update while the system runs: it checks the revocation list, verifies the signed manifest, downloads only what changed, re-verifies the artifacts and stages them. Nothing is switched yet. The switch happens at the next boot, and the boot decides whether it holds: the initramfs checks the anti-rollback counter, reads the write-ahead log, sets up dm-verity and swaps the slots. If the attempt counter runs out the device falls back or enters recovery. Only after enough good boots is the new image promoted.
Where to go next¶
- New here? Start with The model.
- Want the flow end to end? Deployment lifecycle and Boot lifecycle.
- Shipping your own OS on it? Integrating Atom Loops.
- Curious about the guarantees? Trust model.
License¶
GPL-3.0. Source on GitHub.