Coordinator Pattern
Orchestrate multiple agents on a single task and get a unified synthesis.
Some tasks are too complex or too broad for a single agent. The coordinator pattern decomposes a task into subtasks, distributes them to specialized agents, and synthesizes their results.
The coordinator is a supervisory agent. It receives the initial request, analyzes its structure, identifies the required skills, and delegates each subtask to the most qualified agent.
This pattern is particularly powerful for multi-domain tasks: a technical audit requiring both code analysis, security review, and performance evaluation.
Reception
The coordinator receives the request and analyzes its complexity and involved domains.
Decomposition
The task is broken into independent subtasks, each with clear success criteria.
Delegation
Each subtask is assigned to the most qualified agent. Agents work in parallel when subtasks are independent.
Collection
The coordinator retrieves results from each agent, with their confidence levels and notes.
Synthesis
Results are merged into a coherent response. Contradictions are flagged and resolved.
Technical audit
The coordinator delegates code analysis to Max, security review to Eva, and performance testing to a benchmark agent. The final report integrates all three perspectives.
Multi-source research
A research question is distributed to multiple agents with different sources. The coordinator cross-references results and identifies consensus and divergences.
Content generation
A complex article is decomposed: one agent drafts the structure, another the technical content, a third the simplified version. The coordinator assembles everything.
featurePages.coordinator.config
- Available for all EasyClaw v2 installations
- Requires at least 2 configured agents
- The coordinator can be any agent or a dedicated agent