Distributed infrastructure is normal.
Fragmented operations shouldn’t be.
Every cluster answers to its own tab, its own credential, its own endpoint.
The estate is knowable only by asking several people at once.
The same estate, reachable from one place, under one policy.
Centralised operational visibility. Your existing monitoring stays where it is.
Where does Orcastra fit?
Choose the situation closest to your infrastructure.
Your servers. Your cloud. One place to operate it.
Jakarta has a cluster. Surabaya has a cluster. One team watches both.
Sell your cloud, not your vendor’s dashboard.
Many customers, many clusters, one operations team.
Prove who reached what, and on whose authority.
A developer needs a VM, not a three-day ticket.
GPUs are expensive. Know who is using them.
Server down, network gone, SSH with it.
Build a private cloud.
Do not give the infra team a headache.
You own the servers. You want virtual machines and containers to keep running on your own infrastructure.
What you do not want is every node, cluster, project, network and storage pool managed one at a time.
Orcastra becomes the management layer above LXD, Incus or MicroCloud.
Virtual machines, containers, networks, storage, projects and clusters in one place.
Operate many clusters at once, whatever platform each one runs.
Different teams get different access, scoped to the resource.
You can see who did what, on which cluster, and when.
The underlying cluster stays native. Nothing is rewritten.
Switch Orcastra off and the cluster carries on running.
Jakarta has a cluster.
Surabaya has a cluster.
One team still watches both.
Infrastructure grows by location, by business need, or by project.
Eventually one team is opening several dashboards, a VPN, a terminal and a spreadsheet just to know how everything is doing.
Orcastra makes every site legible as a single infrastructure estate.
Which sites are online, and which stopped answering.
How much compute is genuinely still available, per zone.
Which nodes are unreachable, named rather than averaged away.
Which cluster certificate expires soon, before it becomes an outage.
Sell your cloud.
Not your vendor’s dashboard.
A local cloud provider does not have to build an entire control plane from nothing.
Orcastra can be the infrastructure layer behind the cloud service you sell.
Your infrastructure. Your customer. Your brand.
Each customer environment stays separated from the others.
Network, image, profile and storage boundaries inside a cluster.
CPU, memory and storage allocated per customer, validated against the node.
Connect your own portal without handing it admin rights.
Your logo, your brand, your customer experience.
Reaching a customer console is deliberate, recorded and visible to them.
Many customers.
Many clusters.
One operations team.
Customer A runs MicroCloud. Customer B runs LXD. Customer C runs Incus.
Every environment is different. The phone that rings is still yours.
Every customer cluster in one inventory, without a separate login for each.
No standing admin credential sitting in a password manager forever.
What your engineers can do is scoped per customer, not granted globally.
Every session is logged against the customer it touched.
A managed service should leave the customer in control, not require them to hand over every admin credential permanently.
An auditor does not only ask:
“Is the system secure?”
They ask the specific questions, and they expect the answer to already exist.
Modernize the stack.
Keep it understandable.
For organizations starting to build an open-source private cloud, or reducing their dependency on a proprietary virtualization stack.
Orcastra manages the target infrastructure once workloads are running on LXD, Incus or MicroCloud. It does not perform the migration itself.
A developer needs a VM.
Not a three-day ticket.
Give every engineer a project, a shell and a scoped token, and give the platform team one place to see all of it.
The portal provisions against a key that can only reach the projects you named.
The platform team keeps a single inventory across every environment developers touch.
GPUs are expensive.
At minimum, know who is using them.
For an enterprise AI lab, a university AI centre, a GPU private cloud, an ML team or a research infrastructure group.
Orcastra shows which GPUs a node physically has and passes one through. It is not a scheduler and reports no utilisation figure.
Where a node reports no GPU, the picker says so rather than offering an empty list.
Which cards exist, on which node, and how many are still unassigned.
Pick a physical GPU on the target node, or enter a PCI address.
A GPU instance belongs to a project, so the allocation has an owner.
One infrastructure.
Many labs, lecturers, students and projects.
Shared compute, divided into projects with their own boundaries and their own limits.
Server down.
Network gone.
SSH went with it.
Sometimes the dashboard is not enough. You just need a way back into the machine.
Console and terminal run over the cluster’s own channel, so no public IP, no SSH daemon and no working guest network is required.
root@recovery-01:~# mount -o remount,rw / root@recovery-01:~#
The channel belongs to the cluster, not to the guest network.
Nothing has to be installed or running inside the instance first.
Upload an ISO, attach it, and boot to firmware from the same window.
Drag a file onto the terminal and it lands inside the instance.
A DR site checked once a year
is not a DR strategy.
Orcastra does not perform replication and does not replace a DR orchestration system. It gives you visibility and operational access to the recovery infrastructure.
The infrastructure team is four people.
The server count is in the hundreds.
Infrastructure is allowed to grow. The operations team does not have to grow with it.
Let AI read the infrastructure.
Do not hand it root.
Orcastra is not the model. No LLM runs inside the control plane.
It publishes an MCP server, so the assistant your team already uses can read the estate through a key you scoped and can revoke.
An agent starts able to ask what exists, and nothing more.
Each capability is a separate grant, not a role that implies the rest.
An agent call is recorded with its actor and result, like a person’s.
Revoke the key and every agent holding it stops in the same instant.
Which capabilities each situation leans on.
Every capability is available in every deployment. This is only a map of which ones each situation tends to reach for first.
Different industries.
The same operational problem.
Whether you are an enterprise, a cloud provider, an MSP, a university, a government body, a research centre, a telco, a bank or a manufacturer, the problem is usually identical.
Infrastructure is spread out. There are too many tools. Access gets harder to control. And the bigger the estate grows, the harder it becomes to answer a simple question.
What do we have?
Where is it?
Who can reach it?
What is broken?
How much capacity is left?
Orcastra answers all five from one place.
Same infrastructure.
Better operating layer.
Three platforms underneath, one operating model above. Pick per site, manage as one.
Virtual machines and system containers in one platform.
A compact private cloud stack built around LXD, MicroCeph and MicroOVN.
A community-driven system container and virtual machine manager.
Orcastra does not take over
your infrastructure.
LXD stays LXD. Incus stays Incus. MicroCloud stays MicroCloud.
Your workloads remain on the underlying platform, in its own format.
Turn Orcastra off and the cluster keeps running exactly as before.
GPLv3, with a public repository and no contributor licence agreement.
One control plane.
Your infrastructure stays yours.
Manage LXD, Incus and MicroCloud across clusters, sites, teams and customers.