How to use it
- Open the app and pick a module: RAG, tool calling or MCP.
- Press play and watch the agent run, one message at a time.
- Tap the LLM, a tool or the retriever to read what it's doing in plain words.
How it's built
A React and Three.js scene that replays a real agent trace. Each step (prompt, tool request, tool result, answer) is a node in 3D space, and the camera moves between them as the run plays.
Why it matters for businesses
Teams adopting agents lose weeks to confusion about what the model does and what the code does. A shared visual makes onboarding faster and design reviews clearer, because everyone sees the same loop.
In real life
A product manager about to approve an "AI assistant" feature can watch exactly where the model calls a tool and where your code takes over, then ask the right questions about cost and risk.



