← Hub

MurmurSwarm

Agent routing through spreading activation

- concepts - edges - agents
Task
Examples
How this works

No LLM decides which agent handles your task. Instead, your words become seeds in a knowledge graph. Signal spreads through learned connections — the same physics neurons use.

Agents whose expertise domains receive the most activation score highest. The graph is the router.

Instant — no LLM call, pure graph traversal
Emergent — routing arises from structure, not rules
Adaptive — the graph evolves as agents work

Try Simulate to see the full cycle: score → agent works → graph changes → re-score → converge.

Agent Resonance

Type a task to see how the knowledge graph routes it to the right agent.