Once you've done the few sign-ups, this is the whole install. You paste one key; the installer does the rest inside your own Google Cloud, pausing only for your passkey.
Click Sign in with Google on paracoding.ai. That creates your account — no signup form — and establishes the identity your fleet will act as. This is the same account whose Cloud project the fleet deploys into.
Point the installer at a Google Cloud project with billing enabled, and grant a short-lived ops window — a time-boxed Google token that lets setup run as you for the next 12 hours, then expires.
Paste your Anthropic (Claude) API key once. Then the installer enables the needed Google Cloud APIs and creates your Gemini / Vertex AI key for you — because you're already Google-authenticated with billing on, no console hunting required.
One command stands up the whole core in your project: the serverless MCP control plane, the example fleet (Advisor + Infra), the data lake, and the event bus — with least-privilege service accounts wired from the start.
$ paracoding deploy # runs in your project, from the open-source scaffoldOpen the gate, enter the one-time setup password, and register this device with Face ID / a passkey. The setup password then permanently disables — from here on, new devices are added only from an already-trusted one. This passkey is your commit key for everything the fleet stages.
Open the cockpit — on your desktop or your phone — and say hello. Tell Advisor what you're trying to build; it plans the work and queues the fleet. Want the streamed desktop, dashboard, or one-tap Windows VM? Those are premium unlocks you can turn on with a key, including a free trial.
Once you're in, you can ask any agent for help and it walks you through the rest. The GitHub repository — with the deploy runbook and the honest threat model — publishes with the launch.
