FREE · PDF
THE COMMAND CENTER FIELD GUIDE
How I run a fleet of AI agents from one screen and my phone.
One email with the guide. No spam, no list.
ASK TOBY Running now
The Command Center
How I run a fleet of AI agents from one screen and my phone.
Why I built it
A month ago I had 15+ VS Code windows open across two monitors. Every window was a different AI agent on a different computer, and I couldn’t keep straight which window was which machine, which agent, or what it was supposed to be doing.
I dreamed about it. The dream was StarCraft. In a real-time strategy game you never wonder who a unit is. Every unit has a portrait, a job, and a place on the map. You click it, you see its face, you know what it does.
So I built that.
The map
Everything hangs off one main node: the conductor that watches every agent, the brain that remembers what the fleet has done, and the rail that carries every report and question to my phone.
Why the faces matter
Every face is a status light. Waiting holds still on purpose: fourteen agents animating at once dropped my wall to four frames a second.
How one decision travels
- 1
An agent hits something it can’t decide alone.
- 2
It sends one card: what it found, two or three complete options, its pick first, and a safe default: stop and report.
- 3
I tap a button on my phone, or rule on the wall. Both write the same decision record.
- 4
My answer lands back in that agent’s session, and it keeps going.
Three rules I’d keep
One path to me.
Agents never ping me from ten places. Every report and every question is a card on one rail.
Every decision is a card.
If it isn’t on a card with buttons, it isn’t a decision.
Plan first.
Before an agent changes anything, its plan goes on my phone. Until it does, the system won’t let it touch anything.
There are a lot more rules than these. Every one exists because something went wrong first.
What’s next
Coming soon
Pro
Building each piece, step by step: the main node, every page, and how a new agent joins the fleet.
Coming soon
Premium
The hard part: the rules that keep a fleet like this from hurting you, the real failures behind each one, and the templates I use.
No prices yet. This page is where they’ll show up first.
What broke
Every rule in this system exists because something went wrong first.
Every agent has a lead. A failing grade has a face.
- INCIDENT 01
THE BOT THAT ATE MY REPLIES
What happened
two programs read my phone chat at the same time. They didn't crash. They silently stole each other's messages, so my answers vanished.
The rule
exactly ONE process talks to my phone. Everything else goes through it.
- INCIDENT 02
FORTY "NEEDS YOU" CARDS
What happened
the wall flagged an agent "needs you" every time it finished a turn. Forty red flags, almost none of them a real decision. I stopped looking.
The rule
"needs you" means an open decision, counted from the same record as the decisions page. Nothing else gets to cry wolf.
- INCIDENT 03
FOUR FRAMES A SECOND
What happened
fourteen agents were waiting on me at once, all animating, and the wall dropped to four frames a second.
The rule
when an agent waits on me, its face holds still. The label says who's waiting.
- INCIDENT 04
THE WRONG ACCOUNT
What happened
two agents shared one browser. One switched accounts while the other was mid-edit, and one brand's profile got another brand's name.
The rule
one agent per login at a time, and every post re-checks which account is active in the same glance as the Post button. Wrong account? Don't click.
BEFORE YOU GO
KEEP GOING
FREE · PDF
THE COMMAND CENTER FIELD GUIDE
How I run a fleet of AI agents from one screen and my phone.
One email with the guide. No spam, no list.
ALL-ACCESS
GET EVERY PROJECT, AS IT'S BUILT
Every build log, uncut footage, my own files, and every new project, for one monthly price.
FOUNDING PRICE COMING SOON
JOIN THE WAITLIST