One person. A whole company.

Most of the companies that should exist never do. Not because the idea was wrong — because the person who had it could not afford the twenty people it would have taken to find out. For every founder who got to try, there are hundreds who priced it up, did the arithmetic, and went back to work on Monday. Ambition has always been a payroll question.

That is the part that is about to stop being true.

We think he is right, and we think the billion is the least interesting number in that sentence. What matters is that the ceiling everybody has built under is moving, and it will move first for whoever is ready for it.

Being ready is the hard part, and it is not a question of how clever the models get. Companies do not run slowly because the people in them are slow. They run slowly at the joins: the figure retyped by hand out of one system and into another, the handoff that waits until somebody happens to notice it, the status nobody has updated, the thing one person knows and nobody ever wrote down. Not one of those is anybody deciding anything. It is a person being used as wiring between two systems that were never connected, and the company runs at the speed of that wiring.

The manual cell is not somebody's judgement. It is somebody's afternoon, spent doing a job a machine was already doing on either side of them. Those are the humans worth taking out of the loop, and they are the only ones: the person who decides what ships is not a bottleneck in the process, they are the reason there is a process. Take the person out of the retyping, never out of the deciding.

That is the whole argument for agents, and about half an argument for Crewlet, because a pile of agents is not a company. Hand twenty of them the same inbox and what you have built is a group chat: nobody owns the outcome, nothing gets checked, two of them do the same job twice, and the bill arrives at the end of the month with no name on any line of it.

Companies solved that for people a long time ago, with something nobody puts on a slide. Structure. Somebody owns the result. Somebody signs it off. Everyone knows what they are for, who they answer to, and what they are allowed to spend.

Crewlet exists to give that structure to agents. So that a company of one can run like a company of a hundred — and a company of a hundred can stop losing its afternoons to the joins.

Crewlet is an open-source engine for running a company whose staff are AI agents. You describe the company you want — the departments, the roles, who reports to whom, what each seat is responsible for and what it is allowed to spend. Crewlet gives every seat its own agent, with its own name, its own memory and its own budget, and puts it to work where your team already works: chat, the tracker, the code host, the docs.

Change the shape of the company and the company changes with it, while it is running, with every version kept and every change attributed. However many humans there are — one of you, or fifty — they stay in the chart, and in charge.

Most questions are already answered in the documentation. For everything else, here's how to reach us — and if you are building something with it, we would genuinely like to hear about it.