Run AI agents on data you cannot put in the cloud.
Two tracks buy this for two different reasons, and both get the same deployable: Console, running inside your boundary, with aethyr-node carrying device traffic to it.
Regulated enterprise
Your data is not allowed in a shared cloud.
Healthcare, finance, legal, and anyone whose counsel has already said no to a hosted model provider. Console runs inside your boundary, and the evidence an auditor asks for accumulates while it runs.
Defense and government
Your network may not reach the internet at all.
Programs handling CUI and mission data, from a fixed installation to a disconnected edge. No call-home, no telemetry, and post-quantum cryptography in the transport rather than on the roadmap.
What both tracks get.
Four things a hosted model provider cannot give you at any price. Everything else on this page is a variation on where they run.
Agents are principals, not borrowed logins
An agent holds its own DID, its own signing key, and its own privileges, so recurring autonomous work is attributable to the agent rather than to whichever human it was impersonating.
An audit spine an assessor can sample
Every action is hash-chained and signed with ML-DSA-65, re-verified on a nightly schedule against published keys. Tampering, key rotation, and configuration gaps stay distinguishable from each other.
Compliance collected while the system runs
Twenty-four frameworks modeled as code. Automated checks and evidence capture for a growing subset, structured attestation workflows for the controls a human has to assert.
One deployable, inside your boundary
Runtime, API, admin surface, and data layer ship together, with no per-feature SaaS sprawl to review. Private cloud today; on-premises and air-gapped delivery are the roadmap.
How these are built is on the technology page.
Not sure which track you are in?
Tell us the constraint you are working against and we will tell you plainly whether Console clears it today.