Why ComOS Is Not Open Source

Ronald Reynolds · 2026-09-12 · 5 min read

Every time someone who runs their own servers hears about ComOS, they ask the same thing before anything else. Is it open source?

Here is the answer, word for word, the same every time: The discipline artifacts are open source. The platform is not. The node is a sealed, checksummed artifact from a public releases repo that you run on your own hardware; your data lives in your own local database; the exit is your data, on your disk, in a documented format.

In plain terms: the way we work is public. The software itself is closed. You get a sealed package with a number stamped on it that proves nobody has altered it since we sealed it, and you run that package on a machine you own. Your business records live on that machine, in files you can open without us. If you ever leave, you take the files with you, and their layout is written down.

That is a complete answer. But I have come to think the question is standing in for something else. Most business owners who ask about source code will never read a line of it. What they want to know is narrower, and it comes in four parts. Can I see what this thing is doing to my business? Can I leave? What happens to me if you go under? And will you change the deal on me later?

Those four deserve straight answers. A license file is not one. Can I see what it is doing?

Reading source code is one way to check software. It is slow, it goes stale the day you stop reading, and almost nobody who asks for it does it. What business owners actually do is trust a person who read it, or trust nobody.

ComOS gives you a different kind of look inside. The running system answers questions about itself. Ask it what changed last week. Ask it what it is unsure of. Ask it where its own records have gaps. It will tell you, and it will show you the record it is answering from. Your own agent can put those questions to the exact copy you are running, every morning if you like, which is more than a pile of source code has ever done for anyone.

We say the same thing to every skeptic. Have your agent interrogate the live system and check our claims against what it finds. Can I leave?

Your records sit on your own disk. Customers, orders, the ledger, the instructions you have given your agents. All of it. Leaving means keeping what is already yours. (If you would rather keep the records on a server you run somewhere else, that is one line in the configuration. Your call, not ours.)

An exit that lives behind an export button is a promise from the vendor. An exit that is already sitting on your own hard drive needs no one's cooperation. What happens if you disappear?

The package keeps running on your machine. The stamp on it stays valid. There is no subscription clock and no renewal date that switches your business off. Standing still costs nothing. The system charges when money moves, and only then. A vendor who cannot bill you into a corner has very little power over you, and that stays true whether the vendor is thriving or gone. Will the deal change later?

You will never hear "we'll open it up later" from me. Plenty of businesses built on open source that changed its license once they were too deep to leave. The promise is what burned them. A no you can plan around is worth more than a maybe you cannot verify. The real reason

There is a reason underneath all four, and I would rather say it than have it discovered.

From the first day, ComOS has been one human and a fleet of AI agents. No staff, no committee. That is the experiment itself. The project exists to show that one person, working with agents, can build and run an entire commerce operating system with the books kept honest the whole way.

Open source would bring in an organization. A queue of outside changes waiting for review, and contributors who, having done the work, rightly earn a say in what gets built next. That is healthy for a normal software company. It would dissolve this one. The structure would survive a small team. It would not survive a crowd.

So the method is what we publish. The two checks that keep one human honest are public on GitHub, and anyone can run them on their own work. One fires whenever a change is proposed. The other fires whenever anyone says a piece of work is finished, and it holds the claim until the proof is on the table. Nobody gets to skip that second check, and the reason is blunt: whoever misjudged the work will also misjudge whether the work needs checking.

I think publishing code I could not support would be the least trustworthy thing I could do. People who run their own servers punish an abandoned repo harder than a confident no, and they are right to.

We hide nothing about who we are or how we work. What stays closed is the org chart, and the org chart is one person.

The package is public. The stamp is on it. Have your agent check it.