Skip to content
Nash Software Services
NASHSOFTWARE SERVICES
← Writing

I Built an Agent Control Room

My AI agents were getting unruly — silent failures, no visibility. So I built a real-time control room to fix it.

AIAgentsAutomation

My army of minions (agents) were getting a bit unruly. Things were silently failing and I had absolutely no idea which ones, or why. So I built a control room.

A couple of days ago I posted about how I got a Notion board working autonomously — agents picking up tickets, doing the work, updating status. It was great. Until it wasn't.

The problem was visibility. Despite the Notion ticket updates and the terminal logs, I still couldn't tell what was actually going on. Silent failures were slipping through, and trying to manage that through columns and agent instructions alone wasn't cutting it.

So I started building something to pull it all together.

The control room gives me:

  • Agent status at a glance
  • Start/stop from the browser
  • Real-time logging — verbose when I need it, human-readable when I don't
  • An audit trail of events
  • A report viewer of completed work
  • Agents that can raise their hand when something goes wrong — spawn error, recent failure, log anomaly — so I can ask Claude about it with full context, right there in the panel

It's not perfect yet. But it's working, and honestly it's been a fun thing to build alongside everything else.

Visibility isn't a nice-to-have when you're running agents. It's the whole game.

(Want to see the repo? Drop a comment.)

Share

Want practical AI agents built for your business — not demos, not experiments?

See AI & Automation

Related posts