I built a memory layer for AI agents that tracks beliefs over time and handles contradictions. Looking for people to test it.

I've been building something called OMEM and I'm at the point where I need people to actually try it and tell me where it breaks. Most agent memory today is basically a list of facts in a vector store. When two facts conflict, one quietly overwrites the other and the history is gone. That always bothered me, so I built something different. What it does: Tracks what each agent believes over time, not just a static pile of text. Every fact has a state (believed, contradicted, unknown) that the engine works out from the evidence. Handles contradictions instead of hiding them. If two things conflict, it surfaces the conflict instead of picking a winner silently. Keeps provenance. You can ask why something is believed and get the chain that led there. Cross-agent memory. Memory is private to an agent by default, and you choose what to share with a team or the whole project. Semantic recall. It finds relevant memories even when the wording is different from how they were stored. A learning loop. Memories that turn out to be useful get ranked higher over time. It runs locally with no external services. Install is basically pip install and start a server, and there's also a web dashboard if you want to see your agent's memory, conflicts, and the belief graph visually. To be upfront: this is early. It works and it's tested, but it is not polished and it is not production ready. I'm looking for people who find the problem interesting enough to poke at a rough thing and tell me what's wrong, what's missing, or what feels off. Honest criticism is exactly what's useful right now. It's completely free for testers. I'm not selling anything and I'm not looking for customers yet, I just want real people running it against real agents. If you want to try it, message me and I'll send you everything you need to get set up. Takes about a minute to get running. Happy to answer any questions in the comments too.