loopyard

An operator runs the yard

One agent that reads status, dispatches work, and only interrupts you when a human is actually needed.

The same Loopyard session on a phone: An operator runs the yard

Ask where things stand, get a real answer, hand out the next job in the same breath. The rail keeps running work and open questions in view.

Once you have a few projects with a few agents each, the hard part stops being “can the agent do the work” and becomes “who is keeping track of all of it.” That is the operator’s job.

The operator is an agent whose whole world is your Loopyard. It reads every workspace’s status, pulls recent activity, and answers “where do we stand?” with a short summary instead of a wall of logs. Tell it to ship something and it hands the task to the right workspace agent, then gets out of the way.

It is built to leave you alone. Work that finishes quietly stays quiet, and only the things that need a human reach you, in a list with one-tap answers. Overnight breakage gets fixed where it can be and summarized where it cannot.

The operator runs in its own container on your machine like everything else. It has no shell on your host; its power comes from the same control-plane tools every agent uses, so anything it does is visible and auditable in the same UI you are already looking at.

Run it on your machine

git clone github.com/loopyard/loopyard.git
cd loopyard
mix loopyard.setup

Free and open source on GitHub. Follow @loopyardai.