Private infrastructure · Operational

One control room.
Every machine.

Bizzy Command Center turns a scattered Mac-and-Windows fleet into one understandable workspace. It shows which computers are ready, what capacity is available, and starts Codex on the machine that actually owns the project.

  • v0.3working system
  • Mac + Windowshost-native agents
  • Private by designno public control surface

In plain English

It is the page I open instead of walking from computer to computer.

My projects and AI tools do not all live on one machine. Instead of remembering where everything is, opening remote desktops, and rebuilding terminal context by hand, I choose a computer and project from one private page. The Command Center makes the connection, starts or resumes Codex there, and keeps the operational details out of the way.

Fleet control

Your machines, without the noise.

Live health across the fleet, plus one direct path to start Codex where the work lives.

Public studio window

See the kitchen, not the keys.

A delayed, read-only glimpse of aggregate fleet activity. It shows the shape of the work without exposing machines, addresses, ports, models, prompts, projects, or personal data.

BIZZY / PUBLIC ACTIVITY WINDOW PREPARING
Fleetmachines online
Memorycoarse pressure band
Codingcoarse activity signal
ENGINES ON THE FLOOR The first privacy-filtered snapshot has not been published.
  • Private exporterNOT ENABLED

Coarse and delayed by design. This surface cannot control the fleet.

Why it exists

The hardware was already a system. It needed to feel like one.

THE PROBLEM

Useful capacity was scattered.

Different projects, models, and services belong on different computers. That created a daily tax: check each machine, find the right folder, open the right terminal, and remember which session was still alive.

THE ANSWER

Centralize the view, not the work.

A small controller presents one dashboard while host-native agents keep execution on each machine. The interface coordinates the fleet; the files, credentials, tools, and compute stay where they already belong.

The everyday flow

From “where is it?” to working.

The core interaction is deliberately short enough to use from a laptop, tablet, or phone.

  1. 01

    See the fleet

    Check availability, CPU, memory, disk, GPU, uptime, and the health of configured services.

  2. 02

    Pick the right home

    Choose an online machine and a discovered project folder within its approved roots.

  3. 03

    Start with context

    Launch Codex in that folder and optionally send the opening prompt with the session.

  4. 04

    Come back later

    See managed sessions in one place and reattach without reconstructing the setup.

How it fits together

One coordinator. Small agents at the edge.

The browser never needs direct administrative access to each computer. The controller authenticates the person, polls the fleet, and proxies the approved session traffic.

A useful distinction

What it is—and what it is not.

IS

A private operations layer

It makes a personal fleet legible and gives repeatable access to project-aware Codex sessions.

IS NOT

An AI model

Codex and local services do the specialized work. The Command Center coordinates where that work runs.

IS NOT

A public remote shell

Access stays on the private network; service actions are allowlisted, and session launches begin only inside approved project roots. The public artifact is this case study—not a live admin panel.

Design decisions

Power, with guardrails.

Because this interface sits close to real machines, the safety model is part of the product design.

01

Private by default

The dashboard is intended for an encrypted personal network, with no public ingress.

02

Least necessary access

Working directories are bounded and service actions are explicitly allowlisted.

03

Credentials stay server-side

Each host has its own identity; node credentials are not handed to the browser.

04

Operations leave evidence

Selected privileged control-plane actions—such as session creation and stop events—are recorded in a local audit log with common secrets redacted.

Under the hood

A web interface with host-native reach.

The controller is a Python service with a compiled React interface and a local SQLite audit trail. Lightweight Python agents expose only the monitored state and allowed operations needed by the dashboard. Native pseudo-terminals keep command-line behavior correct across macOS and Windows.

Interface
React · TypeScript · xterm.js
Controller
Python · FastAPI · SQLite
Machine bridge
PTY · ConPTY · system telemetry
Private transport
Tailscale · HTTPS · per-node auth

Current state

The useful version exists.

Version 0.3 adds a native, read-only iPhone dashboard and the privacy-filtered public studio window to fleet status, project discovery, managed Codex launch and reattachment, cross-platform terminals, service health, and audit logging. It is an operational personal tool, not a public SaaS product.

NEXT

Make sessions outlive the bridge.

The next architectural step is durable session ownership across service restarts, followed by deeper structured Codex integration and clearer multi-machine handoffs.

That is what the heck Bizzy Command Center is.

Explore the rest of the work