MOOF Knowledge Graph

The MOOF Knowledge Graph is the platform’s central repository of structured knowledge that agents can leverage for informed reasoning. It is both a global knowledge resource (common facts and ontologies that many agents might need) and a local customizable graph (domain-specific data for a particular universe).

Role of the Knowledge Graph:

• Grounding Agents in Facts: LLM-based agents are prone to hallucinating incorrect information. By querying the Knowledge Graph, an agent can verify factual claims or look up data instead of guessing. The Knowledge Graph acts as an authoritative reference library. When combined with the agent’s natural language reasoning, it produces more accurate and context-aware results​.

• Facilitating Multi-Agent Reasoning: In scenarios where multiple agents collaborate, a shared knowledge graph helps synchronize their understanding. One agent’s output can be written as structured data to the graph, and another agent can later read and reason over it. This is akin to a shared blackboard in classical AI systems or a common database in multi-agent systems. This “single source of truth” prevents divergence in the agents’ individual mental models of the world.

• World Modeling: For an agentic universe that simulates a complex system (say an economic simulation or a virtual city), the Knowledge Graph can represent the state of that world – entities (people, places, objects) and their relationships. Agents then act not just on text inputs but on a rich world state. The graph can enable multi-hop reasoning (like Agent A deducing it should talk to Agent B because B is the manager of department that oversees task T, as encoded in the graph relations).

In combination, the Memory Network and Knowledge Graph give MOOF agents something that early frameworks lacked: persistence and knowledge grounding beyond what fits in an LLM prompt. This dramatically enhances the realism and reliability of Agentic Universes created on the platform.