One daemon
Four front doors. One API.
The CLI, the console, Kubernetes and Terraform all talk to the same API, so nothing drifts between them.
- CLI`fabricctl` creates, inspects and scripts everything the API can do.
- Web console92 pages: fleet health, VMs, networks, storage, security and Merion.
- Kubernetes operator`VirtualMachine` resources next to your pods, turned into API calls.
- TerraformA provider is included. Plan your private cloud like any other.
Fabric decides what should exist. FluxVM makes it exist. GuestKit prepares the disk.
What is inside
Everything a private cloud needs.
Compute and network
VM lifecycle and software-defined networking, with a TC/eBPF dataplane that enforces per-VM policy and rate limits.
Network Fabric ›Secure by default
PAM, LDAP and OIDC sign-in, JWT auth, three-tier RBAC, audit export and encryption at rest. Built in, not bolted on.
Security ›Private inference (Beta)
OpenAI-compatible endpoints on the same daemon: models, Maglev-weighted backends and API keys. Single-cluster Beta, not GA.
AI Workloads ›
Merion
Your agent gets a real computer. You keep the keys.
An open workstation for an AI agent you can’t fully trust: a sealed microVM, network rules the host enforces, and approvals only you can give.
- Approve from anywhereYour iPhone or Android phone signs each decision with a key that never leaves it. Slack and Teams get the same request with the host-rendered preview; buttons appear only when no phone signature is required.
- See everything it didEvery action lands in a signed audit trail. Send it to any OpenTelemetry collector; an allowlist keeps file contents, previews and credentials out.
- Propose, then decideThe agent speculates in an isolated copy of the cell. Its changes wait behind an approval, and the agent can never approve.
- Sealed by designA deny-all network policy set by the host, credentials that never enter the cell, and a counter that shows zero outbound connections.
Fleets
Every cluster. One view.
Register your other Fabric daemons and see their hosts, VMs and health together. A host that is down never breaks the list, and maintenance runs on the target host’s own API with your own permissions.
Get started
Five minutes. One server.
Build, install, start the daemon, and create your first VM. Needs a Linux server with KVM.
git clone https://github.com/zyvorai/fabric.git && cd fabric
make build && sudo make install
sudo systemctl enable --now zyvor-fabricd
fabricctl create web-01 --image fedora-41 --cpus 2 --memory 4096 --tenant acmeExpect Console at https://localhost:9095/app
- Linux with KVM
- Ports 9095 (API and UI) and 7788 (FluxVM, localhost)
Honesty
What we don’t claim yet.
- Live migration: the disk-copy path is GA; the native FluxVM transport is preview and full HA cutover is pre-GA.
- AI Workloads is single-cluster Beta, not GA.
- Merion evidence is
software-test. The Slack, Teams, OpenTelemetry, federation and Android features are tested against local fakes, not yet a live workspace, collector, second daemon or device.