RookeryOS documentation

What RookeryOS is, who it is for, and how this documentation is organised: platform, architecture, security, deployment and the stack.

Why RookeryOS exists and how it replaces the SaaS API economy with a browser-first agent runtime: pods, the browser as API, and the human loop.

Product requirements, the architectural blueprint and the phased MVP milestones for RookeryOS, from the first agent pod to the managed platform.

The Kernel + Sidecar pod design: how the two containers talk, what they share, and why the split keeps the agent isolated and observable.

Provisioning, bootstrapping, running and tearing down agent pods, from the request that creates one to the cleanup that reclaims it.

The state-driven knowledge system behind every agent: the Registry, the Librarian, the Reasoner and the HTML renderer, and how documents flow.

Sub-second WebRTC voice for agents with local speech-to-text and text-to-speech: the pipeline, the latency budget and where it runs.

The zero-trust, five-layer security model: the Cookie Vault, namespace isolation, RBAC and encryption, and what each layer refuses by default.

How RookeryOS builds, ships and rolls back: container images, the Kubernetes rollout, secrets and the checks a release must pass before traffic.

The operator console for RookeryOS: provisioning pods, watching health and cost, granting autonomy, and the human-in-the-loop queues.

Every layer of the RookeryOS stack: Kubernetes and GKE, Node and TypeScript, Postgres, Chromium sidecars, Gemini models, WebRTC voice and nginx.