simregistry
Even well-trained AI agents need a reliable representation of the world they operate in.
Simregistry integrates with existing agent frameworks through a simple API or MCP server, allowing agents to query what has happened, record new information, and update the simulated world.
Analyze systems, not just individual agents.
Simregistry makes it possible for developers, researchers and auditors to inspect how agents behave individually and how their interactions shape the wider system.
- Are agents behaving logically or erratically?
- Is the multi-agent system stable, healthy, or degrading?
- Are key metrics moving in the right direction?
Simregistry was developed as part of research into ethical resource allocation and won the KXSB Hackathon in the Ethics, Agency & Social Impact category.
The same infrastructure has since been used in production to maintain and manage the gamestate of online simulations.