Start with the hypervisor. Grow into a private cloud.
Every edition has Kubernetes, the Aurakore AI Gateway, the console, the API and Aurakore AI, with no separate management fee. The editions differ in shared storage, availability and what is built in around them.
In every editionFrom Compute Standard up, nothing to add
Live migration
Preview on Compute Standard
KubernetesPreview
Clusters on platform VMs
Aurakore AI GatewayPreview
Local models first
MicroVMsComing
Picked at deploy
Console and API
No management fee
Aurakore AI
Plain-language answers
Audit ledger
Every change recorded
PreviewACS
The essential hypervisor
Aurakore Compute Standard
Virtual machines on each host's own disks, for smaller clusters of up to three hosts that want a dependable hypervisor and nothing to build around it.
Create, resize, clone and delete VMs; host maintenance
Live migration of running VMs, their local disks moved with them Preview
Bridge networks with an optional VLAN
Console, API and Aurakore AI on the appliance
Hardening checks against a secure baseline on every host
Host and VM metrics sent to the collector you already run
Tamper-evident audit ledger
Everything it comes with
INSTALL AND RUN
One image, every hostOne signed image and a first-boot wizard. A cluster installs in about 20 minutes with no hand stepsPreview
Long-run stabilityRan 92 hours under continuous memory load on test hardwarePreview
Runs disconnectedFully air-gapped. Nothing needs a connection to the outside worldPreview
COMPUTE
Virtual machinesCreate, resize, clone and delete; add and remove disks; host maintenancePreview
Local-disk storageEach host's own empty disks become its VM storage. A disk holding data is refusedPreview
Host isolation detectionHosts tell an isolated host from a failed one by agreeing among themselves, with no shared storagePreview
Virtual TPMA virtual TPM with secure boot firmware, keys held by the cluster; it moves with the VM, and each clone gets its ownPreview
Bridge networksBridge networks with an optional VLANPreview
Overlay networksPrivate networks spanning hosts, isolated by defaultPreview
KubernetesCreate, scale and delete Kubernetes clusters on platform VMs, in every editionPreview
Content libraryBase images checked against a declared checksum, and ISO discsPreview
OPERATE
Console and APIGuided forms in the console, each the same operation as its API twinPreview
Aurakore AIPlain-language answers from live cluster state, on the appliance. Your data stays in the buildingPreview
Aurakore AI GatewayOne governed path from agents and apps to granted tools and models; credentials stay sealed, every call recordedPreview
Hardening and driftHosts checked against the secure baseline; drift fixed only with a second factorPreview
Metrics to your toolsHost and VM metrics in the standard format your collector reads, with a revocable read-only credentialPreview
Performance and capacityHost and VM charts up to 90 days (per VM, CPU only); days-until-full estimates and rightsizingPreview
Alarms and logsBuilt-in checks sent to webhook, syslog, email or SNMP; central log forwarding with searchPreview
Browser VM consoleA VM's screen or serial line in the browser, with a fresh authenticator codePreview
AutomationA command-line client and a Terraform provider built from the published APIPreview
ACCESS AND RECORD
Accounts and rolesLocal accounts with an authenticator app; custom roles from the platform's privilegesPreview
Audit ledgerEvery change recorded in a hash-chained, append-only, tamper-evident ledgerPreview
Single sign-onOpenID Connect, SAML 2.0 or LDAP and Active Directory, with a second factor always requiredPreview
Sign-in limitsEnds after 15 minutes idle or 12 hours; destructive actions still ask for a fresh codePreview
Scoped roles and foldersRoles on a cluster, folder or single VM; folders limit VMs, vCPUs, memory and diskPreview
Generally availableAEC
Shared storage and availability
Aurakore Enterprise Compute
Everything in Compute Standard, plus Aurakore SAN and the availability features that keep workloads running through a host failure.
Everything in Compute Standard, and:
Aurakore SAN, readable through a host failure, rebuilding on its own
Live migration over Aurakore SAN, with no disks to copy
Automatic restart, only after the failed host is confirmed down and fenced
Predictive evacuation from a host showing signs of failure Coming
Automatic balancing, each move recorded and reversible
Snapshots and placement rules (keep together, keep apart, pin)
Everything it comes with
SHARED STORAGE: AURAKORE SAN
Distributed storageOne replica per host, readable through a host failure. Capacity is never metered
Loss of one host's drivesEvery data drive of one host lost: guest writes carried on with no errors, and full redundancy was back within a minute of the drives returning
Drive repairA lost drive's data copied again from its replicas; a returned drive restarted on its own; damaged data repaired, each repair recorded
Rebuild after host lossA lost host's data copied again to the hosts that remain, on clusters with more hosts than copiesPreview
Storage manager takeoverA standby takes over storage management within 30 secondsPreview
Templates and linked clonesA template deployed as many VMs: linked clones, full copies, and a clone made independent
AVAILABILITY
Live migrationRunning VMs move between hosts, confirmed on multi-node hardware
Automatic restartA failed host is confirmed down and fenced, then its VMs restart elsewhere. Never without a confirmed fence
Predictive evacuationVMs moved off a host before it fails, on the host's own health signalsComing
Automatic balancingOff, recommend or automatic. Each move recorded and reversible; switchovers measured at 95 to 222 ms
SnapshotsNative disk snapshots, as of one instant (crash-consistent), and restore
Snapshots with the guest agentThe guest's file systems paused through its agent while a snapshot is takenPreview
Placement recommendationsWhere each VM should run, honouring keep-together and keep-apart rules and host pinning
VM encryptionVMs encrypted in memory on hosts with the processor feature; never placed on, or moved to, a host without itPreview
RECORD
Ledger reviewView the audit ledger and run tamper checks from the console
INSTALL AND RUN
One image, every hostOne signed image and a first-boot wizard. A cluster installs in about 20 minutes with no hand steps
Long-run stabilityRan 92 hours under continuous memory load on test hardware
Runs disconnectedFully air-gapped. Nothing needs a connection to the outside world
COMPUTE
Virtual machinesCreate, resize, clone and delete; add and remove disks; host maintenance
Virtual TPMA virtual TPM with secure boot firmware, keys held by the cluster; it moves with the VM, and each clone gets its ownPreview
Bridge networksBridge networks with an optional VLAN
Overlay networksPrivate networks spanning hosts, isolated by default
KubernetesCreate, scale and delete Kubernetes clusters on platform VMs, in every editionPreview
Content libraryBase images checked against a declared checksum, and ISO discsPreview
OPERATE
Console and APIGuided forms in the console, each the same operation as its API twin
Aurakore AIPlain-language answers from live cluster state, on the appliance. Your data stays in the building
Aurakore AI GatewayOne governed path from agents and apps to granted tools and models; credentials stay sealed, every call recordedPreview
Hardening and driftHosts checked against the secure baseline; drift fixed only with a second factor
Metrics to your toolsHost and VM metrics in the standard format your collector reads, with a revocable read-only credential
Performance and capacityHost and VM charts up to 90 days (per VM, CPU only); days-until-full estimates and rightsizing
Alarms and logsBuilt-in checks sent to webhook, syslog, email or SNMP; central log forwarding with searchPreview
Browser VM consoleA VM's screen or serial line in the browser, with a fresh authenticator codePreview
AutomationA command-line client and a Terraform provider built from the published APIPreview
ACCESS AND RECORD
Accounts and rolesLocal accounts with an authenticator app; custom roles from the platform's privileges
Audit ledgerEvery change recorded in a hash-chained, append-only, tamper-evident ledger
Single sign-onOpenID Connect, SAML 2.0 or LDAP and Active Directory, with a second factor always requiredPreview
Sign-in limitsEnds after 15 minutes idle or 12 hours; destructive actions still ask for a fresh codePreview
Scoped roles and foldersRoles on a cluster, folder or single VM; folders limit VMs, vCPUs, memory and diskPreview
ComingAEC-D
A hypervisor for VDI
Aurakore Enterprise Compute for Desktops
The Enterprise Compute hypervisor, built and licensed for virtual desktops. It runs under the connection broker and display protocol you already use; it is not a desktop broker itself. Desktops run in pools, beside the few infrastructure VMs a desktop estate needs.
Desktop pools: pooled, persistent and session-host desktops, built from one image held once on Aurakore SAN
Desktops kept ready ahead of the morning logon, recycled and updated in rolling batches
Up to 16 infrastructure VMs per cluster, each with a declared role: broker, gateway, profile storage, directory and the like
Desktop-only by design: general-purpose VMs belong on Enterprise Compute
Aurakore SAN, live migration, automatic restart and automatic balancing; predictive evacuation is coming
Aurakore Perimeter available as an add-on
Everything it comes with
DESKTOP POOLS
Desktop poolsA pool kept at its size from one template. A released desktop is deleted and a clean one built in its placeComing
Persistent desktopsOne desktop per person, kept across sessions. Another person is refused itComing
Image updatesA new template version rolls through idle desktops; one in use is never touched until it is releasedComing
Scale inA smaller size removes idle desktops only. Desktops in use stayComing
Desktop-only fenceA VM is a pool member or one of up to 16 infrastructure VMs (brokers, gateways and the like)Coming
Templates and linked clonesPools are built from linked clones of a template; full copies and independent clones tooComing
Windows desktop imagesWindows desktop images in pools. The hardware runs used Linux desktopsComing
Everything in Aurakore Enterprise Compute, included.
ComingACP
The complete private cloud
Aurakore Cloud & AI Platform
Everything in Enterprise Compute and Enterprise Compute for Desktops, with Aurakore Perimeter and Aurakore Resilience built in, on hardware you own and fully disconnected if you need it.
Aurakore SAN, live migration, automatic restart behind fencing and automatic balancing
Desktop pools for the connection broker you already run
Aurakore Perimeter: microsegmentation, intrusion prevention, a web application firewall, load balancing and gateways
Aurakore Resilience: restore points, locked off-site copies, replication and failover to a second site
GPU passthrough, GPU partitions and model serving are planned for this edition
Everything it comes with
GPUS AND AI SERVICES
GPU passthroughA whole GPU for one VMComing
GPU partitionsHardware slices of one GPU per tenant, with quotas and chargebackComing
Model serving and catalogYour approved models behind an industry standard endpointComing
Everything in Aurakore Enterprise Compute and Enterprise Compute for Desktops, with Aurakore Perimeter and Aurakore Resilience built in.
Generally available
Aurakore Perimeter
Network security for every workload, as an add-on to any edition or on its own. Every rule runs in shadow first, so you see what it would drop before it drops anything.
Microsegmentation between workload categories, kept in force through live migration and restart
Inline intrusion prevention, fail-open or fail-closed, and a web application firewall that replays traffic before it blocks
Load balancing that rides through a host loss, and one service name answered from whichever site is healthy
Gateways, floating addresses, port forwards and a gateway firewall
Outbound rules by service name, and flow analytics: who talks to whom
A change that loosens security asks for a second factor
Everything it comes with
SEGMENTATION AND INSPECTION
MicrosegmentationAllow and deny rules between workload categories, tried in shadow first, then enforced. Held through agent, leader and fabric restarts
Policy follows the workloadRules stay in force through live migration and automatic restart on another host
Intrusion detectionTraffic between workloads checked against your rule sets, with alerts per hostPreview
Inline intrusion preventionBlocks matching traffic. You choose fail-open or fail-closed; a stopped engine is reported
Web application firewallDetect and block modes, with a replay of captured traffic before blocking
Application-layer rulesMethod, path and host rules on load-balanced servicesPreview
NETWORKS AND TRAFFIC
Load balancingHTTP traffic spread across healthy backends; a failed backend is removed and returns, and service continues through a host loss
TLS terminationAn HTTP load balancer holds the certificate and decrypts traffic for its backendsPreview
Gateways and published addressesOutbound access through a gateway, a floating address for one workload, or a port forwarded to it
Global load balancingOne service name answered from whichever site is healthy
Name-based outbound rulesAllow outbound traffic to named services; an address follows its name as the name changes
Gateway firewallEdge rules, shadow before enforce; the platform's own addresses stay reachable
Site-to-site linksEncrypted tunnels, route exchange with your routers, and networks stretched between sitesPreview
VISIBILITY AND AI WORKLOADS
Flow analyticsWho talks to whom, by workload, port and direction, including traffic leaving the cluster
Suggested rulesRules proposed from recorded traffic, for an operator to createPreview
Point-in-time reachabilityAnswers whether X could reach Y at time T, with the rule and the record behind the answerPreview
AI workload blueprintsReady-made policies for AI agents and model servers that never loosen without approvalPreview
Agent egress triggersAn alarm, then optional containment, when a workload's outbound traffic changes sharplyPreview
Coming
Aurakore Resilience
Disaster recovery as an add-on, between Aurakore sites you own, through your own storage. The sites never connect directly.
Scheduled restore points per group of workloads, with retention
Encrypted copies in your S3-compatible storage, locked against deletion, restorable as a new VM
Replication to a second site, with a recovery point down to 5 minutes
Planned failover with no data loss, the source fenced first; failback when the site returns
Unplanned failover only once a person confirms it with a second factor, and runbooks that recover many workloads in order
Everything it comes with
PROTECTION
Scheduled restore pointsSnapshot schedules per group of workloads, run automatically or on demand, with retentionComing
Off-site copiesEncrypted copies in your S3-compatible storage, locked against deletion, restorable as a new VMComing
Replication between sitesAsynchronous copies to a second site through your own storage, recovery point down to 5 minutes. The sites never connect directlyComing
Near-synchronous replicationSeconds of data loss between nearby sites, on encrypted storage traffic onlyComing
RECOVERY
Planned failoverMove workloads to the other site with no data loss. The source is stopped and fenced firstComing
Unplanned failoverStart workloads at the second site when the first is down, once a person confirms it with a second factorComing
FailbackReturn workloads to their original site once it is backComing
RunbooksOrdered recovery of many workloads in one action; a failover can name one to start by itselfComing
Failover testsA copy started on an isolated network with no way in or out, then removed, without touching productionComing
Network mappingEach network mapped to its counterpart at the second site, with each workload's network identity keptComing
Readiness scoreOne score per site for protection coverage, backup health and recovery readinessComing
Alarm to runbookAn administrator binds an alarm to a recovery runbook, started or offered for approval when it firesComing