Hi Everyone! I built Compass because I think a big part of intelligence comes from team dynamics. I believe we did not get to where we are as a civilization only because we have a powerful brain. Instead, we got better and better because different people bring different experiences, challenge each other, debate assumptions, and eventually converge on something clearer than any one person started with. Most AI products collapse that dynamic into one assistant, one context, single experience. Compass explores a different path: what happens when multiple AI agents, each with its own role and perspective, can work inside the same project? This is a research preview built on Copilotz, our open-source framework for building AI agent harnesses. Today Compass has four agents: North keeps the work aligned with the goal. East explores alternatives and non-obvious paths. South verifies claims and separates evidence from assumptions. West stress-tests risks and constraints. I would love feedback from builders, researchers, and anyone who has felt the limits of single-assistant workflows.

Copilotz Compass

A research preview for team-of-agents AI work

A research preview for team-of-agents AI work