One-person trading desk

By the TradeAgentic engineering team·Published 30 August 2026·Updated 30 August 2026

One person, a desk's worth of jobs

A trading desk is a set of continuous responsibilities, and a person is not continuous. The question is not whether one person can make good decisions. It is whether one person can make them at the same standard on a Tuesday afternoon and at three on a Thursday morning, every week, without missing one.

The answer is no, which is the honest case for automating the desk rather than the trader. What follows is which jobs have to be covered before one person can run this, and which never leave you.

What has to be covered before you can run it alone?

Protection that does not depend on you being awake

Protective orders resting at the brokerage rather than inside the software. If the stop exists only in a running process, then a crash, a reboot or a closed laptop lid removes it and nothing tells you.

Recovery without a human

A power cut at four in the morning should not mean an unattended open position and no software running. The system should come back, resume the mode it was in, and reconcile positions and protection before opening anything new.

Reconciliation you do not perform

Rebuilding realised profit and fees from the brokerage's own record, automatically, before new risk is opened. Skipped by most products, and the reason software and reality drift apart quietly.

A record you can actually read

With no team to ask, the decision record is your only route to understanding what happened. If you cannot follow it, you will not read it, and an unread record is the same as no record.

What still lands on you?

The size decision

Whatever the paper record showed, how much real capital goes in is yours. Size it so total loss is annoying rather than damaging, which is how you buy the right to keep going after a bad month.

The stop decision

Not the stop on a position, the stop on the whole enterprise. Write down in advance what would make you halt, because deciding it during a drawdown is when people decide worst.

The reading

A cadence you hold to, looking at refusals as well as trades. This is the job that quietly lapses after a few good months, and it is the one that makes the difference between running a desk and hoping.

What does a solo operator get wrong?

Treating a quiet period as proof

Every autonomous system eventually gets ignored because it has been fine for a while. Nothing has been proven; you have simply stopped looking.

Intervening during a drawdown

The reduction in an approach that stopped paying is the system working. Treating every decline as a fault reintroduces exactly the discretion the design removed.

Running it on a working laptop

The machine you close at the end of the day is the wrong machine. Sleep, updates and reboots at the wrong moment are the most common way one person interrupts their own system.

Is one person enough?

For the unattended jobs, yes, if the software covers them

Reconciliation, verified protection and clean restart are what decide it. Entry logic is the part every vendor demonstrates and the part that matters least to whether you can leave it alone.

For accountability, there is no one else

Autonomy concentrates responsibility rather than removing it. You decided to run it, at that size, inside those limits, and that is the whole of the exposure.

For the discipline, it is entirely on you

No committee is forcing patience through the paper period, and nobody else will notice if you stop reading the record.