MoAI ADK - Multi-agent Orchestration Interface
Purpose
Orchestrate multiple AI agents for complex tasks using the MoAI framework.
Core Concepts
Agent Roles
- •Planner: Breaks down complex tasks
- •Executor: Performs specific actions
- •Reviewer: Validates outputs
- •Integrator: Combines results
Workflow Orchestration
- •Task decomposition
- •Agent assignment
- •Parallel execution
- •Result aggregation
- •Quality verification
MoAI Tool Integration
Tools Available
- •Task dispatch
- •Context sharing
- •Result aggregation
- •Conflict resolution
Usage Patterns
Pattern 1: Sequential Pipeline
code
Input → Agent A → Agent B → Agent C → Output
Pattern 2: Parallel Processing
code
┌→ Agent A →┐
Input → ├→ Agent B →┼→ Integrator → Output
└→ Agent C →┘
Pattern 3: Review Loop
code
Input → Executor → Reviewer → (Approved → Output)
└→ (Rejected → Executor)
Best Practices
- •Clear Interfaces: Define inputs/outputs for each agent
- •Context Management: Share relevant context between agents
- •Error Handling: Plan for agent failures
- •Progress Tracking: Monitor multi-agent workflows
- •Result Verification: Validate final outputs
OpenCode Integration
Works seamlessly with OpenCode's subagent system:
- •Use
subagenttool for agent dispatch - •Leverage
skilltool for capability loading - •Monitor with
todotool for task tracking
Usage
Activate for:
- •Complex feature development
- •Multi-step refactoring
- •Cross-domain tasks
- •Quality assurance workflows