Skip to main content
ZWRM Agents are persistent coding agent VMs running on isolated Firecracker microVMs. Connect your preferred AI coding tool (Claude, Codex, etc.) to a full VM with a home directory volume that survives across sessions. Agents have full access to the ZWRM platform — the zwrm CLI is pre-installed and authenticated. Your agent can build, deploy, scale, and manage applications directly from within its VM. Ship code from ideation to production without leaving the agent session.

​
Quick start

zwrm agent claude
This creates (or reconnects to) a Claude coding agent VM and drops you into an SSH session.

​
How it works

  1. VM creation — A Firecracker VM is provisioned with your chosen agent type.
  2. Persistent home — A volume is attached to the VM’s home directory, so your files, config, and history persist across sessions.
  3. Platform access — The zwrm CLI is available inside the VM with your credentials, so the agent can deploy apps, manage secrets, create databases, and more.
  4. SSH session — You’re connected via SSH with the agent’s default command running.

​
What agents can do

From inside a coding agent VM, you (or the agent) can:
  • Deploy apps — zwrm deploy to build and ship directly from the agent
  • Manage secrets — zwrm secrets set to configure environment variables
  • Create databases — zwrm postgres create to spin up Postgres instances
  • Scale services — zwrm scale to adjust replicas
  • View logs — zwrm logs -f to stream application output
  • SSH into apps — zwrm ssh to debug running VMs

​
Supported agent types

TypeDescription
claudeClaude Code agent
codexCodex agent

​
Named instances

By default, each agent type gets one instance. Use named instances to run multiple environments: 
zwrm agent claude my-project
zwrm agent claude another-project
Each named instance has its own VM and persistent volume.