Security you do not have to take on trust
Protection is part of how the platform is built, not an option you switch on — and every claim on this page is something we test.
What keeps your business safe
In plain words, without the technical detail.
Your data stays where you decide
Set the countries and regions where your data may live. Any placement that breaks the rule is refused, with the reason, and cannot be forced.
A password is not enough
Approving a payout can require a second factor or a signature, and the most critical actions several people. Each person sees only their region and only the fields they need.
History cannot be quietly rewritten
Every action and every refusal is recorded — who, what, when, why. Editing or deleting the record is detected, and any past figure can be re-checked.
Stolen disks are useless
Disks are encrypted and the key is never stored on the server. Connections already use post-quantum key exchange by default.
Updates without fear
Every update is signed and checked. A new version is tried first; if something goes wrong, the server returns to the previous one by itself.
Nothing foreign inside
Our own operating system and no third-party libraries in the working code: a flaw in someone else's component cannot reach you through the back door.
Who can do what with your servers
Management of every node is locked down by design.
No remote login to abuse
Servers are managed only over an encrypted channel where both sides check each other's keys. There is no remote shell.
Separate roles for separate jobs
A monitoring role can watch but change nothing; an unlock role can only unlock disks.
Disks watched around the clock
Data is verified in the background, and suspect disk areas are quarantined automatically until an operator clears them.
5 hard power cuts, nothing lost
Power was cut 5 times during writes on a dedicated AMD EPYC server with NVMe (October 2026): all 248 acknowledged writes were intact, none lost, none corrupted.
Built for audit and data-localisation rules
Placement rules by country, a complete history, and results your auditor can reproduce.
We state only what we test. For example: a 6-hour run under load in an x86-64 virtual machine with 4 vCPU (October 2026): 17 update cycles — 8 confirmed, 9 rolled back automatically — 35 reboots and about 407,000 HTTP operations. Every confirmed piece of data was intact and memory use stayed flat.
See it on your own business
Book a demo: we show the platform on your processes and agree where your data will live — in our cloud or on your own nodes.