The onpit stack

We invented none of it, on purpose. Every component is proven, open or European, and runs entirely on your own hardware. Our work is choosing, fitting together, securing and keeping it current.

Layer by layer

running today on the onpit cluster   in progress

Applications

  • Maildocker-mailserverOpen sourcerunning
  • Files, calendars, contactsNextcloudNextcloud GmbH, Stuttgartin progress
  • Editing documents togetherCollabora OnlineOpen sourcein progress
  • Video callsJitsi MeetOpen sourcein progress
  • ChatMatrix with ElementOpen source, open protocolin progress
  • ERP, CRM, accountingOdooOdoo S.A., Belgiumin progress

AI brain

  • Language modelsopen models, e.g. from Mistral AIMistral AI, Parisin progress
  • Running models locallyOllama, Open WebUIOpen sourcein progress

Identity & security

  • One login for everythingZitadelZitadel AG, St. Gallenrunning
  • PasswordsVaultwardenOpen sourcein progress
  • Security monitoringWazuhOpen sourcein progress

Network & access

  • Encrypted network between all devicesNetBirdNetBird GmbH, Berlinrunning
  • Access from outside, certificatesTraefikTraefik Labs, Lyonrunning

Cluster & data

  • Several computers as one systemk3sSUSE, Nurembergrunning
  • Storage mirrored across nodesLonghornSUSE, Nurembergrunning
  • Databases with automatic failoverPostgreSQL with CloudNativePGOpen sourcerunning

Hardware & storage

  • Compute nodesRaspberry Pi 5, mini PCs or serversARM64 and x86-64running
  • Separate storage server, provides storage over NFS and S3OpenMediaVault on DebianOpen source, Volker Theile, Germanyrunning
  • Operating system (storage today, nodes with the next rebuild)DebianOpen source, communityrunning
  • Checksummed data storageOpenZFSOpen sourcerunning
  • Backup storage (S3)GarageDeuxfleurs, Francerunning

How we choose

A component only goes into the stack if it meets all five points.

  • It already exists and is proven. We don't rebuild what already exists in good shape.
  • It's open source or European, ideally both.
  • It runs entirely on your own hardware, with no forced cloud and no secret phoning home.
  • It's secure and data-frugal by default.
  • It's replaceable. No lock-in, not even to onpit.

onpit, Google Workspace and Microsoft 365

The same jobs, three very different answers to the question of who owns the data.

Comparison of onpit with Google Workspace and Microsoft 365
QuestiononpitGoogle WorkspaceMicrosoft 365
Where is the data?On your own hardware, on your own networkWith GoogleWith Microsoft
Which law applies?EU lawUS law too (CLOUD Act)US law too (CLOUD Act)
Mail and calendardocker-mailserver, NextcloudGmail, CalendarOutlook, Exchange
FilesNextcloudGoogle DriveOneDrive, SharePoint
Shared documentsCollabora OnlineGoogle DocsOffice Online
Video and chatJitsi, MatrixMeet, ChatTeams
One login for everythingZitadelGoogle accountEntra ID
ERP, CRM, accountingOdoonot includedDynamics 365, separate
AI assistantYour own AI brain, localGemini, in the cloudCopilot, in the cloud
Security monitoringWazuhseparateSentinel, separate
CostHardware once, service optionalper user per monthper user per month
Switching providerAny time, open formatshard workhard work