MAScope

MAScope, developed by researchers, is a security framework that protects multi-agent AI systems by reconstructing and monitoring cross-agent semantic flows to detect execution-aware attacks.

Two-hop subgraph: this entity, every entity it directly relates to, and every entity those neighbors relate to. Drag a node, scroll to zoom, click to inspect — or click any neighbor and re-center the atlas there.