One of the ways I am trying to stay close to what is happening in AI agents outside of work is by building small things in public.

I started with a setup that lets me code from my phone by SSH-ing into a Linux server and using Codex there. That setup has been simple, fast, and surprisingly effective. It means I can refine specs, experiment with AI agents from anywhere without waiting until I am back at my laptop.

The more I use different agent setups, the more I start to feel like that the operational view of all tooling is fragmented. It feels like there should be a simple unified interface for monitoring agents across platforms.

That is the idea behind Agent Ops.

Agent Ops is a small side project I am building to stay close to the pace of development in AI agents while exploring a problem that feels increasingly practical. The first version is intentionally narrow: ingest events, store current state plus history, and show a clean mobile-friendly view of what each agent is doing right now.

The goal is not to build every integration on day one. The goal is to make the first useful thing work.

I am building it in the open here:

https://github.com/osewadare/agent-ops