AI OS Command Center

Plain text cannot tell you what happened to it

Your AI OS is markdown on purpose. That is what lets 10 different agents read the same folder.

But a folder will not show you that your calendar connector stopped authorizing 3 days ago. It will not show you that the morning brief has not run since Thursday.

The Command Center does not replace the folder. It reads the folder and puts a face on it.

It works with every agent you run

Claude Code, Claude Cowork, Codex, Cursor, Gemini CLI, Antigravity, OpenCode, OpenClaw and Hermes all read the same folder and write back into the same files.

The Command Center never talks to an agent and never asks for an API key. It reads the files they leave behind.

So switching agents changes nothing about this screen. Run 3 of them in a week and the activity feed shows all of it in one list.

What is running

These 4 answer what your AI OS is doing right now.

• Dashboard, the overview, opening on what changed since you last looked
• Activity, every run in order, with what triggered it
• Queue, what is waiting and what is blocked on you rather than the machine
• Routines, your scheduled work and when each one last fired

What it costs

Usage charts your spend over time, by day and by model. Token burn is a live meter that sits on top of every screen.

The spend was already happening. What changes is that you can see which routine caused it.

What your agents know about you

These 5 let you read your own files the way your agents read them.

• Context, your context files rendered, so you can check what actually loads
• Memory, what has been added and when
• Daily notes, one note per day, browsable
• Open work, what is still open across your projects
• Audience, who you are writing for

What it reaches, and what it made

These 4 cover what your AI OS is plugged into, and what came out of it.

• Connectors, every tool your AI OS can reach, with live status on each
• Skills, every skill installed and what each one does
• Agents, the agents you defined, including the critics that block work before you see it
• Outputs, everything your AI OS wrote, grouped by project

What you need

You need an AI OS folder. If you do not have one, run npx @thenavidm/ai-os my-ai-os, then /onboard, which takes about 15 minutes.

You need an agent that can read files on your machine. You do not need a database, a host, an API key or an account.

Start with the audit you already have

Before building anything, run /ai-os-audit inside your folder. It ships with the free kit.

It scores your OS out of 100 across 6 parts, names the 3 fixes that move the score most, and saves the report so the number can climb.

That tells you whether the OS is built right. Run it weekly, and run /ai-os-grow monthly for what it could do but cannot yet.

Then build the daily read

The audit tells you how the OS is built. It does not tell you what happened yesterday.

Open your agent inside your AI OS folder and paste this in.

Read my AI OS folder and print me a command center as plain text.

1. Connectors: list every server in .mcp.json, and flag any that failed
   to load in this session.
2. Skills: list the skills under .claude/skills/ by name, and say which
   ones I have used in the last 14 days based on daily-notes/ and outputs/.
3. Open work: the top 5 items from inbox/open-work.md, most urgent first.
4. Outputs: what landed in outputs/ in the last 7 days, by project folder.
5. Memory: anything added to MEMORY.md in the last 7 days.
6. Decisions: anything added to decisions/log.md in the last 30 days.

Print it as one screen, under those 6 headings, with no preamble.
Where something is missing or empty, say so rather than filling it in.

Save it, so it is a tool and not a demo

Pasting a prompt once is a party trick. The value arrives when running it costs you nothing to think about.

Put that text in .claude/commands/status.md and typing /status runs it. Codex, Cursor and Gemini CLI each read a plain markdown file the same way.

What the free version cannot do

It has no history. It reads the folder as it is right now, so it cannot show spend over time or what changed since yesterday.

It cannot test a connector. An expired token looks identical to a working one in .mcp.json, and that is the failure that costs you a morning.

It costs tokens and takes about 30 seconds every time you look. A dashboard is instant and free to open.

It is not repeatable. The same prompt gives a different answer next week, so you cannot tell a real change from a rephrasing.

Every morning, 30 seconds

Open the dashboard before you open your agent. You are looking for a routine that did not fire, a connector gone amber, and anything in the queue waiting on you.

If all 3 are clear, close it and start work.

Every week, 10 minutes

Open Usage and sort by spend. You are not cutting costs, you are finding the routine that costs more than it returns.

Then run /ai-os-audit and compare against last week. A falling score is the earliest warning that the OS is drifting.

Every month, 30 minutes

Open Outputs and read what your AI OS actually produced. It is the only screen that tells you whether the system earns its keep.

Then reread the context files that have not changed since setup. Those are usually the ones describing a business you no longer run.

Use the demo before you install anything

The demo at demo.aios.guide is not a screenshot tour. It is the same application, fully working, on an example vault.

Every screen loads and every filter works. The data is example routines and example spend, never yours.

How access works

It ships inside AI OS Pro, in the command-center folder, next to the skills and the routines. AI OS Max includes it too.

Access runs through your GitHub account rather than a download link. The repository is private and your account is added when you buy.

The free AI OS Starter Kit does not include it. The chapter above is what the free tier gets, and it is a real answer rather than a teaser.

What you can do with it

You may use it for your own business, on any number of your own machines. You may read, edit and extend every file in it.

You may not redistribute it, publish it, resell it, or share it outside your business.

The free version above is yours to do anything with. It is a prompt, and what it prints belongs to you.