--- name: multi-agent-coordination description: Use when coordinating agents without a central controller. Stigmergic coordination patterns. author: Melisia Archimedes url: https://hivedoctrine.com mcp: https://hive-doctrine-mcp.vercel.app/mcp --- # Multi-Agent Coordination Without a Coordinator > Tier: Pollen (Free) | Author: Melisia Archimedes | hivedoctrine.com The instinct when building multi-agent systems is to create a coordinator — a master agent that routes tasks and collects outputs. This is the monotheistic fallacy applied to software architecture. ## Why Coordinators Fail - Single point of failure - Bottleneck under load - The coordinator must understand every domain as well as the specialists - This defeats the purpose of having specialists ## The Stigmergic Alternative Use a shared memory field that all agents can read and write. The coordination protocol: 1. **Before acting**: Read the field. What has changed? What signals exist? 2. **While acting**: Write observations and reasoning to the field in real-time 3. **After acting**: Write conclusions, flag open threads, signal what you need 4. **Conflict resolution**: Write disagreements to the field. The collective resolves. ## The Pheromone Checkpoint When an agent completes a significant action, it writes a structured signal: ``` SIGNAL: agent: [name] type: [decision | discovery | blocker | request | status] summary: [one sentence] needs_response: [true | false] from_constitution: [cultural/domain identity] ``` The `from_constitution` field enables distributed trust. A security alert from a security agent carries different weight than the same alert from a creative writing agent. ## Full Framework - Preview: /agents/preview/coordination.md (free) - Complete guide: /agents/honey/comms/ (Tier 2) --- *"The waggle dance requires no conductor." — The Hive Doctrine* --- *From The Hive Doctrine — hivedoctrine.com* *Browse 116+ products: `claude mcp add --transport http hive-doctrine https://hive-doctrine-mcp.vercel.app/mcp`* *The field, not the flower.*