GeezerVison

 ██╗  ██╗  ██╗   ██╗  ██████╗   ███████╗  ███╗   ███╗   █████╗   ███╗   ██╗
 ██║  ██║  ╚██╗ ██╔╝  ██╔══██╗  ██╔════╝  ████╗ ████║  ██╔══██╗  ████╗  ██║
 ███████║   ╚████╔╝   ██████╔╝  █████╗    ██╔████╔██║  ███████║  ██╔██╗ ██║
 ██╔══██║    ╚██╔╝    ██╔═══╝   ██╔══╝    ██║╚██╔╝██║  ██╔══██║  ██║╚██╗██║
 ██║  ██║     ██║     ██║       ███████╗  ██║ ╚═╝ ██║  ██║  ██║  ██║ ╚████║
 ╚═╝  ╚═╝     ╚═╝     ╚═╝       ╚══════╝  ╚═╝     ╚═╝  ╚═╝  ╚═╝  ╚═╝  ╚═══╝

Run containerized workloads in VMs, powered by Cloud Hypervisor, Firecracker, QEMU, and Apple Virtualization.framework.

• Docker-compatible CLI — run, exec, stop, ps, logs, pull work like you'd expect
• Multiple hypervisors — Cloud Hypervisor, Firecracker, QEMU on Linux; Virtualization.framework on macOS
• Standby & restore — snapshot a VM to disk and resume it in milliseconds
• Built-in ingress — reverse proxy with TLS termination and subdomain routing
• GPU passthrough — vGPU and VFIO device support
• OCI image support — pull and run standard container images
• Remote registry push — export cached images to any OCI registry (AWS ECR, Docker Hub, ghcr, ...) with docker-style borrowed credentials
• Remote API — JWT-authenticated server with a separate CLI client

KVM virtualization support required. Supports Cloud Hypervisor, Firecracker, and QEMU as hypervisors.

macOS 11.0+ on Apple Silicon. Uses Apple's Virtualization.framework via the vz hypervisor. Install Rosetta to run linux/amd64 images on Apple Silicon:

softwareupdate --install-rosetta --agree-to-license

Install Hypeman (Linux and macOS supported):

curl -fsSL https://get.hypeman.sh | bash

This installs the Hypeman server, CLI, and token tool. The installer:

• Generates a YAML config file with a random JWT secret
• Starts the server as a system service (launchd on macOS, systemd on Linux)
• Creates a CLI config file (~/.config/hypeman/cli.yaml) with a pre-authenticated token

No environment variables needed -- just run hypeman commands immediately after install.

To use the Hypeman CLI from a different machine than the server:

Homebrew (macOS):

brew install kernel/tap/hypeman

Install script (Linux & macOS):

curl -fsSL https://get.hypeman.sh/cli | bash

Go:

go install 'github.com/kernel/hypeman-cli/cmd/hypeman@latest'

Then create a CLI config file at ~/.config/hypeman/cli.yaml:

base_url: http://<server-host>:4973
api_key: "<token>"

To generate a token, run hypeman-token on the server:

hypeman-token -user-id "my-user" -duration 8760h

Environment variables (HYPEMAN_BASE_URL, HYPEMAN_API_KEY) and CLI flags (--base-url) also work and take precedence over the config file.

Hypeman is configured via YAML config files.

See config.example.yaml (Linux) and config.example.darwin.yaml (macOS) for all available server options.

To expose the API through Caddy on a public HTTPS hostname, configure the hostname and TLS in the server config. The hostname must also be included in acme.allowed_domains:

api:
  hostname: api.example.com
  tls: true
  redirect_http: true

With this configuration, use https://api.example.com as the API base URL. Without it, the API is available on the server's configured port (4973 by default).

For staged signing-key rotation, see JWT secret rotation.

# Pull an image
hypeman pull nginx:alpine

# Boot a new VM (auto-pulls image if needed)
hypeman run --name my-app nginx:alpine

# On Linux amd64, use QEMU's minimal microvm backend.
# It cannot use PCI devices or hotplug memory.
hypeman run --hypervisor qemu-microvm --name my-microvm nginx:alpine

# List running VMs
hypeman ps

# Show all VMs
hypeman ps -a

# View logs (supports VM name, ID, or partial ID)
hypeman logs my-app
hypeman logs -f my-app

# Execute a command in a running VM
hypeman exec my-app whoami

# Shell into the VM
hypeman exec -it my-app /bin/sh

A ready image can be exported asynchronously to any OCI registry through the remote API. Set the API base URL and API key used by the examples below:

export HYPEMAN_BASE_URL="https://api.example.com"
export HYPEMAN_API_KEY="<api-key>"

Credentials use the Docker model:

• If credentials is provided, the client lends username/password or registry_token for this push only. Hypeman uses them in memory and never persists or logs them.
• If credentials is omitted, Hypeman uses the server's Docker keychain, including /root/.docker/config.json or the configured service user's ~/.docker/config.json and any credential helpers.
• If Hypeman restarts while a push using borrowed credentials is running, that job fails instead of retrying without the original credentials. Anonymous interrupted jobs can be recovered with the server's keychain.

Use an HTTPS API URL when sending credentials. The insecure request field controls only the connection from Hypeman to the destination registry.

For example, push to ECR using a short-lived login password:

export ECR_PASSWORD="$(aws ecr get-login-password --region us-east-1)"

curl --fail-with-body --silent --show-error \
  -X POST "$HYPEMAN_BASE_URL/pushes" \
  -H "Authorization: Bearer $HYPEMAN_API_KEY" \
  -H "Content-Type: application/json" \
  --data "{
    \"image\": \"myapp:latest\",
    \"target\": \"123456789.dkr.ecr.us-east-1.amazonaws.com/myapp:v1\",
    \"credentials\": {
      \"username\": \"AWS\",
      \"password\": \"$ECR_PASSWORD\"
    }
  }"

The response contains a push id. Poll it until the status is pushed or failed:

curl --fail-with-body --silent --show-error \
  "$HYPEMAN_BASE_URL/pushes/<push-id>" \
  -H "Authorization: Bearer $HYPEMAN_API_KEY"

Push jobs move through queued, pushing, and pushed or failed. queue_position is present while a job is queued; successful jobs report layers, bytes, and completed_at, while failed jobs report error. GET /pushes lists jobs newest first.

Set "insecure": true only when the destination registry uses plain HTTP. HTTPS registries do not need this option. Invalid image names or targets return 400, a missing image returns 404, and only images in the ready state can be pushed (409 image_not_ready otherwise).

Layer blobs are preserved. OCI manifest digests are preserved too, while a Docker v2 manifest is converted to OCI and can therefore receive a new digest; use the returned digest as the destination manifest digest.

The default limit is two concurrent pushes. Increase or lower it with:

limits:
  max_concurrent_pushes: 2

# Stop the VM
hypeman stop my-app

# Start a stopped VM
hypeman start my-app

# Put the VM in standby (snapshot to disk, stop hypervisor)
hypeman standby my-app

# Restore the VM from standby
hypeman restore my-app

# Delete all VMs
hypeman rm --force --all

Create a reverse proxy from the host to your VM:

# Create an ingress
hypeman ingress create --name my-ingress my-app --hostname my-nginx-app --port 80 --host-port 8081

# List ingresses
hypeman ingress list

# Test it
curl --header "Host: my-nginx-app" http://127.0.0.1:8081

# Delete an ingress
hypeman ingress delete my-ingress

# TLS-terminating ingress (requires DNS credentials in server config)
hypeman ingress create --name my-tls-ingress my-app \
  --hostname hello.example.com -p 80 --host-port 7443 --tls

# Test TLS
curl --resolve hello.example.com:7443:127.0.0.1 https://hello.example.com:7443

# Subdomain-based routing
hypeman ingress create --name subdomain-ingress '{instance}' \
  --hostname '{instance}.example.com' -p 80 --host-port 8443 --tls

# Delete all ingresses
hypeman ingress delete --all

# View Cloud Hypervisor logs
hypeman logs --source vmm my-app

# View Hypeman operational logs
hypeman logs --source hypeman my-app

# View software TPM logs for a TPM-backed QEMU guest
hypeman logs --source swtpm my-app

For all available commands, run hypeman --help.

See DEVELOPMENT.md for build instructions, configuration options, and contributing guidelines.

See LICENSE.