The operating layer for
autonomous fleets

Podomy builds hived, a partition-native runtime for machine fleets that cannot stay connected. Deploy workloads, keep operating through segmentation, and reconcile state when nodes reconnect.

What is hived?

Kubernetes assumes connectedness. hived assumes partition. Each node carries the control logic it needs to act locally.

Local admission

Each node decides what can run inside the bubble it can see.

Append-only journal

Truth is a local durable event log. State, peers, and workloads are views.

Sync & reconcile

Reconnected nodes exchange missing events and merge deterministically.

Workload contracts

Workloads declare multiplicity, authority, freshness, and merge behavior.

Executor plugins

Containerd runs first. Other executors can support heterogeneous fleets later.

Honest observability

hived classifies everything as known, unknown, stale, conflicted, confirmed, or unconfirmed.

Why this exists

The cloud operating model is coming to machines, but real machines do not stay connected.

Mining

Underground and remote fleets operate through RF gaps where downtime is expensive.

Maritime & subsea

AUVs dive, lose comms, surface, and sync. Partition is normal.

Defense & contested robotics

Mobile assets need software that works without a reliable cloud or network.

Licensing

hived is AGPL. Commercial licenses are available for companies that cannot open-source their modifications.

AGPL

Free

  • Full runtime
  • Local API & CLI
  • All workload classes
  • Sync & reconciliation engine
  • Must open-source modifications
Get started

Get in touch

We are looking for design partners building software for fleets that cannot stay connected.