Spatial agent monitor
Agents Land
Agent activity is easier to follow when it has a place. Agents Land turns coding sessions into a persistent world that can be watched without reading every terminal.
- My role
- Creator & product designer
- The work
- Spatial agent monitor
The problem
A list of sessions can tell you what exists, but it gives little sense of where attention is needed. Sharing that view introduces another problem: useful visibility should not require exposing the contents of the work.
What I built
I designed a spatial monitor in which people have lands, projects become districts, and agents inhabit the world. Work, review, and blocked states give the scene meaning. A local companion connects outward to report supported activity from Codex and Claude Code. Public visitors see aliases and states; prompts, transcripts, tool inputs, and credentials remain on the owner's machine.
The result
Agents Land is available as a public preview, with live lands people can visit and GitHub-based sign-in for creating their own. It explores a different interface for the same coordination problem behind Erich OS and Rider: make activity understandable at a glance, then let a person decide when to look closer.