Launchd BETA
Product Compute Features Docs Request access

Every machine you run.
One tree.

Launchd is a console for compute — cloud regions, colo racks, and the box under your desk, browsed and managed the way you already browse a filesystem. Folders, breadcrumbs, drag and drop, right-click. Nothing to relearn.

Private beta. We're letting teams in a few at a time.

The Launchd console showing a region tree in the left sidebar — us-east · ord1 expanded into production and staging folders — and a staging folder open in the main pane listing eight compute instances with their status, specification, address, and image. A details panel on the right shows the selected instance, stg-api-02, with its usage graphs and tags.

It isn’t a dashboard. It’s a directory.

Every provider we’ve used organizes compute as a flat list you filter and search. Launchd organizes it as a tree, because that’s the shape it actually has: regions contain folders, folders contain instances, instances have children of their own — volumes, snapshots, network interfaces.

Select a region in the sidebar and its folders expand inline, the way a file manager works and a web console never does. Drag an instance from staging into production and it moves — no export, no redeploy, no ticket. Right-click anything for the full action set. The breadcrumb above the list always tells you exactly where you are, because at 2 a.m. you will have forgotten.

  • TREERegions, folders, and instances nest the way your org chart does, not the way the API happened to shape the response.
  • D&DDrag between folders and regions. Multi-select with shift-click, same as your desktop.
  • ⌘KA command palette and full keyboard navigation — the mouse is optional, not assumed.

Your hardware is a region too.

Launchd doesn’t care whether a machine lives in a datacenter we run or a rack in your office. Point the agent at any SSH-reachable box — bare metal, a home-lab server, a colo cage you’ve had since 2019 — and it shows up in the tree next to your cloud regions. Same console, same audit log, same drag-and-drop.

That GPU box under the stairs stops being a special case you SSH into manually and start forgetting about. It becomes a region with a name, a folder structure, and a console tab, like everything else.

  • REACHIf we can SSH into it, we can bring it into the tree — bare metal, a home-lab box, a colo cage, whatever it is.
  • LEAVENo lock-in. Pull the agent off and the box goes back to being just a box.
  • SAMESame audit log, same console, same drag-and-drop as any region we manage ourselves.
The Launchd console console tab open on an instance named stg-gpu-dev, a 1x RTX A6000 GPU box on the own-rack compute provider. A live audited terminal session shows nvidia-smi output and a training script logging loss values, with a connected status and round-trip time shown in the footer.

The hypervisor holds the keys. Not the guest.

Most clouds bake a key into the image at boot and call it done. Need to add a teammate's key later, or cut off someone who just left? You're SSHing in to hand-edit authorized_keys yourself — assuming you still can.

Launchd syncs keys at the hypervisor, not inside the guest OS. Add a key to your account and it's pushed into every VM that should have it — instantly, no redeploy, no logging in by hand. Revoke one and access is gone at the host boundary. The guest doesn't get a vote, and it never even sees the attempt.

  • SYNCAdd a key once and it lands in every VM that should have it, the moment you save it — not the next time someone happens to reboot.
  • ENFORCEAccess control lives at the hypervisor. A guest that's locked up, compromised, or just unresponsive doesn't get a say in who can still reach it.
  • ROTATEPull a key in one place and it's gone everywhere at once — not a per-box chore you'll half-finish.

Three more things worth watching move.

The rest of the console is listed below. These are the parts that are easier to show than to explain.

Virtual machines

Create, start, stop, reboot, resize, and reinstall from base images, with cloud-init baked in — the status dot in the tree flips the moment it's actually true, not when a cache catches up.

Resize & snapshot

Point-in-time root-disk snapshots, plus live resize and reinstall, on the same volume whether it's on our storage or the disk in your rack. Restore, or purge after your retention window.

Networks

Private VPCs with real security groups, enforced at the network layer — not an API-level suggestion. VXLAN overlay means two instances on the same network reach each other across physical hosts without you configuring that.

Built for everything you run.

Compute, containers, networking, storage, and the access control to tie it together — all in one tree, under one account.

Compute

Virtual machines with an in-browser console built in, so opening one doesn't mean finding your SSH client first. VMs, container apps, and static sites all live in the same folders — not three separate dashboards pretending to be one product.

Containers

Deploy a container directly, no cluster to provision first. Point a domain at it, give it the secrets and the persistent storage it needs, and expose whatever ports it actually uses.

Networking

Private networks with firewalls that are actually enforced, not just checked by an API. Every network gets its own router handling NAT, DNS, and a WireGuard VPN in — and a floating IP whenever you need one public.

Storage

Object storage for the things that don't belong on a disk, block volumes for the things that do, and a library of base images to start any of it from.

Identity & access

Teams, folders, and roles that are either admin or member — nothing in between to misconfigure. Invite-only signup, scoped API tokens, and one vault for every SSH key across your fleet.

Platform & operations

An audit log of who did what, webhooks and live event streams for anyone building their own dashboard on top, and per-team quotas that keep one project from quietly eating the whole rack.

Beta access is by invite. Ask for one.

We're letting teams in a few at a time while we finish building this. Tell us what you're running and where, and we'll reach out when there's room.