multi-agent-orchestrator from smithery.ai

📋 Multi-Agent Orchestrator = Competitive Parallel Execution

Core Principle: Launch N CLI agents (Claude Code, Codex, Gemini) with identical task → Compare self-evaluations → Declare winner based on measurable success criteria. 🔴 CRITICAL: NEVER MOCK DATA! Try multiple approaches to get real data; if all fail, stop and document attempts.

Multi-Agent Workflow Structure (Continuant - TD):

graph TD
    Task[Task File] --> Claude[Claude Code Workspace]
    Task --> Codex[Codex CLI Workspace]
    Task --> Gemini[Gemini Workspace]

    Claude --> CPlan[01_plan_claude_code.md]
    Claude --> CResults[90_results_claude_code.md]
    Claude --> CArtifacts[All Artifacts]

    Codex --> DPlan[01_plan_codex.md]
    Codex --> DResults[90_results_codex.md]