< insights

8 min read

One-person team: why and how I built my agent system

I work independently. From talking to clients to deployment, everything is on me. I keep going this way because I've managed to build a system that lets me work at this pace and quality.

But at some point, 24 hours stopped being enough. On one side a client project is progressing, on the other my own product is waiting. When I switch to frontend I forget the backend, bugs pop up while writing tests, scope creeps while fixing bugs. You can do everything, but you can't do everything at the same time.

There's a limit to working faster. I needed to change how I work.


What I built

I built my own agent orchestration. When a project comes in, here's how it works:

First, planning happens. Requirements, tasks, dependencies, priorities get laid out. Then these tasks are distributed to specialized agents. One handles frontend, one handles backend, one handles the database, one writes tests. An orchestrator coordinates all of them, sequences them, and ensures dependent tasks progress in the right order.

This system runs autonomously — I'm not saying it's unreliable, but I don't fully trust it either. I review every piece of code written. I check every test run. The agents do the work, the approval is mine. The responsibility for what's delivered to the client is mine, I can't delegate that to a system.

There's no "AI does everything" story here. There are checkpoints, and I'm at every checkpoint. The agents handle the work in between: code, tests, file structure, linting, refactoring. I make architectural decisions, control quality, and talk to clients.


What this changed

Work that used to take me a week alone, I can now ship in two days. While I'm reviewing one thing, the next task is already being prepared.

The quality of code I wrote when tired would drop. Now every piece of code is written to the same standard. Linting, test coverage, consistent structure. I just look at "is this correct, is this sufficient." I've also become able to run multiple projects simultaneously.

"I'm just one person" is no longer a disadvantage. I don't tell clients "I don't have a team." I have a team, my way of working is different. There's planning, code review, QA, a deploy pipeline. Agents take on these roles instead of humans, with final approval staying with me.


Can everyone do this

I didn't build this system overnight. I learned through trial and error when to trust the agent and when to step in myself. Experience makes the difference here — if you don't know what can go wrong, you can't place the checkpoints right.

But someone who can design it carefully and trusts their know-how can do this. As a solo developer, at some point you either build a team or build your own team. I chose the latter.