OriginTrail is a decentralized knowledge graph (DKG) network that provides a verifiable shared memory layer for AI agents. It encodes data into Knowledge Assets, recording facts, data provenance, and semantic meaning, and supports knowledge publishing, querying/traversal, and trusted sharing. Users can publish and discover knowledge in a privacy-preserving manner by running nodes or using network interfaces, while blockchain ensures tamper-proof and traceable data. Its neuro-symbolic approach combines the structural reasoning of symbolic AI with the pattern recognition of neural AI, grounding AI outputs in verifiable knowledge rather than probabilistic guesses. Knowledge Assets can be organized into Paranets for domain-specific knowledge management.