A complete parking system — entry, exit, passes, monthlies, rates, payments — open source and free to run, on cameras you can order from Amazon. No ticket machines. No per-space fees. No vendor holding your gates hostage.
Early access to Open Parking AI opens on the first of January, 2027.
Until early access opens
No technical knowledge required. It plays on a loop.
Every cost in a parking garage today is the cost of a ticket. The paper. The machine that prints it. The person who drives out to refill it. The call when it jams at seven on a Friday. The exit you give away when someone loses it. The station that reads it.
Take away the ticket and that entire cost structure goes with it. The systems that added cameras kept the ticket as their fallback — which is exactly why nobody's bill ever went down.
The car is the ticket. It always was.
A camera sees the vehicle on the way in and again on the way out, and knows it's the same car. No plate required — plates are useful when they're readable and irrelevant when they're not. Everything else in the system is built on that one fact.
No system to purchase, no licence per lane, no monthly per space. Free to run, for as many garages as you like.
Any camera that streams over a network cable. Buy them online, mount them, pull a wire. No proprietary hardware to be locked into.
No ticket stock, no dispensers to refill, no pay stations, no printer to jam. The moving parts are what break, and there aren't any.
An honest "I don't know" instead of a confident wrong answer. It will never open a gate on a guess, and it will never quietly charge the wrong car.
Dirty plates, paper tags, no front plate, glare, snow, bad angles. The whole vehicle is the credential, so an unreadable plate is not an incident.
The code is public under a licence built to stay that way. If we disappear tomorrow, you still have everything you're running.
A plate reader does one thing: it reads the characters on a plate. When that plate is clean, square to the camera and well lit, it is good at it. The rest of the time it is guessing — and this industry built its entire access layer on top of that guess.
What breaks a plate read: dirt, snow, glare, a tow bar, a bike rack, a paper temporary tag, a dealer frame, stacked characters, a sharp approach angle. And in about half of US states, the plain fact that there is no plate on the front of the car at all.
Those are from 2012 — the last time anyone published read rates with a stated sample. Cameras have improved since. Nobody has published a number in fourteen years, which tells you something on its own.
We don't start with the plate. We start with the car.
The whole vehicle is photographed on the way in — shape, make, model, colour, the dent in the rear quarter, the sticker in the back window, the rack on the roof — and photographed again on the way out. The question is never what does the plate say. It is is this the same car.
The plate is one signal among several. When it reads, it helps. When it doesn't, nothing special happens, because nothing depended on it.
And every answer carries a confidence score. Below the line it refuses instead of guessing, because a refusal is something you can handle and a confident wrong answer is a charge on somebody else's card.
It works best exactly where plate reading falls apart. That isn't a side benefit — it was the brief.
Somebody answers that intercom today. At seven on a Friday with the machine jammed and six cars backed up, somebody is on the phone or in a car driving over. That is not a technology cost. It's a payroll line, and it runs every hour the garage is open.
The newer systems didn't remove it either. They put a conversation engine on every lane — and a conversation engine bills you every time a driver opens their mouth. Your costs now rise with your traffic, which is exactly backwards.
Most cars should get through without anyone saying a word.
So the structured work happens first: who is this, are they allowed, what does it cost, open the barrier. Most cars clear with no conversation at all, because there was never anything to discuss. The expensive part only wakes up when something is genuinely wrong — and then it goes to a person once, with the picture and the facts already on the table.
It also watches the things nobody is watching. A barrier told to open that didn't move. A card reader that stopped answering. Today you learn about those when a queue forms or the day's money doesn't add up. Here you know inside a minute, and so does the owner.
Cheap by design, not by accident. A system that costs you more the busier you get is a system you will eventually switch off.
Take the whole system, or take one piece and run it against the equipment you already have. Each one is a separate system with one way in.
Identifies the car from ordinary cameras and says how sure it is. The part that replaces plate reading.
Runs the lane. Loop fires, camera captures, decision made, barrier opens, and the far loops confirm the car actually crossed.
The voice and the screen at the lane. Answers the intercom, watches for malfunctions, calls a person only when something is genuinely wrong.
Answers one question: does this vehicle have access, at this lane, right now — and if not, why not.
What a stay costs. One place decides price, so the number on the screen and the number on the card always match.
Your monthly parkers. Who is on the books, what they pay, when they are charged, what happens when a card fails.
Where drivers register, swap a car, pull a receipt and manage a monthly. Nobody calls your office.
Not software. Which hardware to buy, how to install it, how to wire it. Public, same as the code.
Every module answers on one contract, so anything can talk to it — ours, yours, or somebody else's. Nothing in this stack assumes our platform is on the other end of the line.
Not everything has to be published. If you want to connect your own system without putting how it works in public, that's a conversation, not a problem. Get in touch.
I started in the parking industry in 1998, as an overnight cashier in a self-park garage. Three months later they moved me to valet.
In 2004, on my birthday, instead of going out partying I sat down and opened my own company, AVPMi. By then I had run every kind of property you can name — hotels, hospitals, airports, restaurants. As they say, there must be a better way. That is why I opened it.
Then I could not install a single location for more than a year.
After that I designed every single location myself, customized it to what they actually needed, and trained every single employee, day and night. Until I sold the company in 2016.
When AI showed up I looked at it and said: these guys are making the same mistakes I tried and solved twenty years ago. So I opened 72Knots.ai. The only goal was to reach the AI companies and teach them my method.
Human or AI. It works the same exact way.
I started building Valet Parking AI as the showcase. It became a product quickly. Then my dream system, BellDesk AI — that became the second showcase, and the second product. Then Locker, the ultimate solution for a ticketless process.
And now the one I am considering my masterpiece: Open Parking AI. Open source code, all of it.
And I am still waiting for the AI companies ;)
I am not a code writer. As a matter of fact, I have never written a line of code, and I had never even seen one, all my life.
Same company, same method, same foundation underneath. Each of these is a working system, not a plan.
A complete valet operation — the lane, the runners, the tickets, the payments and the reports. It and Open Parking AI run as one system on one identity for the vehicle.
valetparking.ai →The bell desk, run the same way: bags, deliveries, guest requests and the people moving them, without a clipboard or a walkie-talkie in sight.
belldeskai.com →Storage and handoff — bags, coats, skis, dry cleaning. Anything checked in and handed back, on the same claim the car already has.
uselocker.app →The parent company, and the method every one of these was built with. The work itself is published there, episode by episode.
72knots.ai →Parking software is expensive because it has been allowed to be. Closed systems, proprietary hardware, a licence for every lane, a fee on every space, and a contract measured in years. The technology moved on. The business model did not.
So we are publishing the whole thing. The code, the wiring, where to buy the hardware, how to install it. Not a trial, not a community edition, not a teaser — the working system, under a licence that means nobody can take it back, close it up and sell it to you again.
You should own the software that controls your gates and your money.
And we are not a surveillance company. We do not track people or watch public space. The system does two things: collect the parking fee, and control who goes in and out. When that's done, the job is finished.
The code is public today. Not a preview, not a snapshot we'll publish on launch day — the working repositories, as they are, while they're being written.
Every commit, every review, and every mistake caught before it shipped is in the history. The reader that read a licence plate off pure static. The camera check that said "yes, there's a car" while looking at a black screen. They're all in there, with the fix beside them.
That's deliberate. A finished repository tells you nothing about how it was made. This one shows the whole method — what got asked, what came back wrong, and what caught it.
This costs you almost nothing, and of the little we do make, a part goes straight back out to people who need it more than we do. That isn't a campaign and there's no badge on the website for it. It's just what happens to the money.
When this saves you something — and it will — do the same with a piece of it. Any amount, any cause you believe in, paid straight to them and never through us. Tell us where it went if you want us to know.
Nobody is counting and nobody is checking. We would just rather this ended up being a thing people did something with than a thing people saved money on.
Those are ours. Yours is yours — pick anyone you like, and we'll never say a word about it.
The first of January is the day the doors open — not the day everything is finished. Early access means the system as it stands, with our hands still on it, in the hands of people who want it early enough to shape it.
If you are waiting for a polished launch, wait. If you would rather be running parking software nobody has ever charged you for, and you don't mind being early, this is for you.
We'll be in touch before the first of January.
One address, we don't share it, and an unsubscribe link on everything we send.
Integrations, partnerships, a question the site didn't answer. It reaches a person, and you'll get a straight answer — including "no" or "not yet" when that's the honest one.
We read everything. You'll hear back.
Or reach us directly.