Running multiple Ai Agents on the same machine, with a Dashboard: Best configuration?

I’m not an expert in this area yet, but I’m trying to figure out the best way to build a system where I can run multiple AI agents simultaneously on the same physical machine, with each agent having its own autonomous GUI, mouse, keyboard/input, applications, and working environment.

My initial idea was to use one PC for developing and configuring the agents, and then have a separate, dedicated machine that actually runs them. Ideally, I’d like the system to have a central dashboard where I can see what all the agents are doing, check their reports, and monitor their status, progress, performance, errors, etc.

I’ve heard some people have a structure of agents like the manager, the supervisor, an agent doing some specific work, then another like quality control.

So that for example, if one agent fails to complete a task, the supervisor could notify me, restart the agent, investigate what went wrong, or potentially even diagnose and fix the problem itself.

The overall goal is basically to build a small “AI office team” where different agents have different responsibilities and can autonomously handle various tasks. However, I want the system to remain visual, so I can actually see what each agent is doing through its own GUI rather than having everything happen invisibly in the background.

The tasks I have in mind are mostly simple, repetitive, computer-based tasks rather than GPU-intensive workloads like video editing, 3D rendering, or movie production.
Because of that, I’m also trying to determine what the cheapest hardware would be capable of running several agents reliably.

I’m also wondering whether it makes more sense to run the agents locally rather than relying heavily on cloud-based AI services. If I have several agents running continuously, the cost of using remote/cloud models could become quite expensive. My thinking is that if I build the infrastructure locally, I could also potentially replace or upgrade the underlying AI models in the future as better GPT or other models become available.

So, what I’m mainly trying to understand is:

  • What operating system would make the most sense?

  • What software/frameworks should I use to build and manage the agents?

  • What should the overall architecture look like?

  • Should I use virtual machines, containers, separate user sessions, or something else to give each agent its own isolated GUI?

  • How should the agents communicate with each other?

  • How should the supervisor/manager agent be implemented?
    etc

Curious to hear from you guys