Three months of building. You do it in a day.
A self-hosted, offline-capable command station for your business: a 3D galaxy of your whole operation, a live agent board, and a voice HUD where a six-agent AI team introduces itself and takes orders. You get the exact prompts, the repos and the templates — no research, minimal tokens.
What you'll have running tonight.
Three static HTML files that talk to each other on your own machine. Open them across your monitors and the station is live.
Command Brain
A 3D force-directed galaxy of your business — every department a glowing node you can drag, zoom and fly through.
Agent Board
A live KPI dashboard — vitals bars, count-ups and a per-agent plate that re-themes as each teammate takes the floor.
JARVIS HUD
An arc-reactor core with your agent roster and a roll call — each agent introduces itself out loud in the free browser voice.
Everything you need to build it in a day.
Three months of building and refining, packaged as a single day's work. Open the guide, paste the prompts, drop in your own data — done by tonight.
- ✓The 1-day build guideStep-by-step, in build order: three screens wired together and talking by this evening.
- ✓The exact copy-paste promptsThe prompts I actually use to generate each screen, the roster and the data model — no prompt engineering on your end.
- ✓The 3 GitHub libraries + install stepsthree, 3d-force-graph and three-spritetext — with both CDN and npm commands and pinned versions.
- ✓The 6-agent roster templateJARVIS, CARI, TOM, SCOUT, OLIE and JULS as ready-made role archetypes you rename and re-role.
- ✓The blank CONTEXT / company-data templateOne file with fill-in placeholders that drives every screen — your departments, vitals, skills and nodes.
- ✓The exact starter files, copy-paste readyThe HUD, board and galaxy shells, the agent roster, the company-data template and the tiny Python build step that inlines everything into one offline .html.
- ✓The Windows launcherA .bat + PowerShell launcher that opens all three screens across your monitors in one double-click.
- ✓Lifetime updatesEvery improvement shipped to the station is yours — free, like the rest of the library.
No payment, no subscription — twelve lessons, four knowledge checks, a final exam and a certificate of completion. If you arrive from an old checkout link, it simply sends you to the coffee jar.
Meet your AI team.
Six role archetypes come built in. They roll-call themselves on the HUD, each in its own voice — and you rename and re-role any of them to match how your business actually runs.
JARVIS Command
Your manager. Reads the whole station, routes work to the right teammate and briefs you first.
CARI Support
Customer support. Handles questions, tickets and follow-ups in your voice, never generic.
TOM Developer
Your builder. Ships the code, wires the screens and fixes what breaks.
SCOUT Research
Your researcher. Digs up what you need, compares options and hands back the short version.
OLIE Content
Content operations. Turns ideas into posts, scripts and copy on a schedule.
JULS Growth
Growth and business. Watches the numbers, spots the leaks and pushes the next move.
Three months vs one day.
I'm not pretending this was quick to figure out. It wasn't. The point is that you skip the figuring-out.
- ·Picking the right 3 libraries out of a dozen dead ends, and the versions that actually play together.
- ·Getting three screens to talk on one origin without a server.
- ·Wrestling free browser voice into wake-word + reply without the mic hearing itself.
- ·Designing one data file that drives a 3D graph, a dashboard and a HUD at once.
- ·Burning a lot of AI tokens on prompts that didn't land the first time.
- ✓The 3 libraries, pinned and installable, with the wiring that works.
- ✓The BroadcastChannel setup already done — paste and the screens sync.
- ✓The voice layer solved, including the anti-echo trick.
- ✓One CONTEXT template that lights up all three screens.
- ✓The prompts that already work — minimal tokens, no trial and error.
Straight answers.
Do I need to know how to code?
No. You paste the prompts into the AI you already use, and it writes the screens. The CONTEXT file is fill-in-the-blank, and the launcher is a double-click. If you can follow a recipe and paste text, you can build this.
Does it cost tokens to run?
Near zero. Once it is built, the station is a static HTML file that runs offline in your browser — the 3D galaxy, the board and the voice all run locally with no API calls. You spend a small amount of tokens once, during the one-day build, generating the screens.
Is my data private?
Yes. Everything runs on your own machine. Your CONTEXT company data is inlined into a local file and never leaves your computer — the whole thing works with the internet off. Nothing is sent to me or anyone else.
What if I get stuck?
You get the community. Post where you are stuck, see how other builders wired theirs, and borrow what is already working. The guide is written for someone building it for the first time, so most snags are already answered.
Your AI command station, running tonight.
The exact prompts, the repos, the templates. Three months of work, done in a day — on your machine, with your data.