Back to blog
EngineeringDate unavailable· min read

Neurologist's Window: Seeing Agent Memory in Three Dimensions

We shipped a 3D force-directed graph that turns agent memory into something you can see, navigate, and understand. Here's why we built it and what it reveals about how autonomous agents think.

Memory is the foundation of agent intelligence. Without it, every conversation starts from zero. Every task forgets the last. Every decision ignores context.

Strug Recall has always given agents persistent memory—facts stored across scopes, confidence-weighted retrieval, spreading activation between related concepts. But until now, that memory was invisible. A black box you could query but never see.

What We Built

The Neurologist's Window is a 3D force-directed graph visualization inside Strug Recall. It renders agent memory as a living neural network: nodes represent memory entries, edges show relationships, and physics simulation organizes the structure in real time.

Built with Three.js and react-force-graph, the visualization lets you:

• Navigate memory clusters by scope (global, role-specific, task-specific)
• See confidence and recency encoded in node size and color
• Trace spreading activation paths between related concepts
• Click any node to inspect the full memory entry
• Watch the graph evolve as agents write new memories

Why It Matters

When you give agents memory, you need to see what they're learning. Not just for debugging—though the Neurologist's Window makes that dramatically easier—but for trust. You need to know what patterns are forming, what assumptions are crystallizing, and where knowledge gaps exist.

The 3D graph makes memory legible. You can see dense clusters where agents have deep context (like our product naming conventions or tech stack), sparse regions where they're still learning, and surprising connections where one insight triggered another.

It's also just beautiful. Watching the graph reorganize itself as new memories write feels like watching neurons fire. Which is exactly what it is.

What's Next

This is the first version. The foundation is solid, but there's more to explore:

• Time-based playback—scrub through memory history to see how agent understanding evolved
• Semantic zoom—collapse clusters into conceptual nodes, expand for detail
• Multi-agent overlays—compare memory graphs across different agent roles
• Query visualization—highlight the exact spreading activation path for any memory retrieval

The Neurologist's Window is live in Strug Recall now. If you're running Strug Works, open the memory browser and click the 3D view toggle. Watch your agents think.