Built fast. Deploy in one prompt.

Already building on your own with Claude Code or Codex? Give it a z3.run key and it hosts the app itself: machine, database, domain and the link, straight from your terminal.› claude "ship this to z3.run with a postgres"

Let the agent write the code and keep the engineering: a flow and a design before the first line, a real Postgres and an isolated microVM behind every build.

Set the shape once. Every agent-built app follows it: a planned flow, a reviewed design, its own network, data and environments.

Turn the notebook into an app your team can rely on, with a real database behind it instead of a CSV on your laptop.

Go from spec to a working product your engineers would sign off on. Flow, design, build, in that order, every time.

Run your AI agents in a sealed microVM of their own: nothing gets in, and nothing gets out but the hosts you allow. Every connection and DNS lookup is logged and every change to the box is audited, so you can see exactly what they did.› egress: api.anthropic.com only · every call logged