Ralv Alpha

Everything runs on a single 3D canvas you can zoom and pan. Zoom out and you see every agent working at once. This is the core idea borrowed from real-time strategy games, where large numbers of units stay manageable because the view scales with them instead of forcing you to scroll through a list that grows longer with every task you hand off. For you, it means adding a fifth or tenth agent doesn't bury the others under a scrolling list you'll never finish reading.

When an agent needs a decision, Ralv signals it on the canvas rather than hiding the request inside a log file you'd have to open and read line by line to notice. Pan across the view and you can tell at a glance which agent is blocked and which one is still working. That cuts the back-and-forth of opening each session to check status. No more guessing.

Clicking an agent opens its work so you can read what it's doing without leaving the interface. You stay in the same mental space instead of switching windows. As we understand it, this is where merge review happens, letting you compare changes from a single agent before you accept them. One click, full context.

Ralv doesn't replace your agents. It sits on top of tools you already pay for. The site notes you can use existing Claude Code or Codex subscriptions with it. So there's no second bill for model access. Your tokens, your provider, and your existing subscriptions all stay exactly the same as before you installed it.

Chat windows reset your context when you move on. A spatial layout keeps position and status visible, so returning to a task takes a glance, not a re-read. That matters more as the number of agents grows past what a list can hold cleanly. Position becomes memory.