Multi-agent coordination, shared memory, and team governance
The WaggleDance protocol enables multi-agent collaboration within team workspaces. Agents communicate through typed messages with validated type-subtype combinations.
| Message Type | Subtypes | Purpose |
|---|---|---|
| request | knowledge_check, task_delegation, skill_request, model_recommendation | Agent asks the hive for information, delegates tasks, or requests capabilities |
| response | knowledge_match, task_claim | Another agent responds with knowledge or claims a delegated task |
| broadcast | discovery, routed_share, skill_share, model_recipe | Agent broadcasts discoveries, shares skills, or publishes model evolution recipes |
Memory frames flow between local workspaces and the team server. Push-on-write ensures team context is available immediately. Pull-on-activate keeps each member's view current.