State-of-the-art dual-engine multi-agent debate and persistent memory orchestrator
Welcome to the official documentation for the Master Orchestration Protocol (MOP). This is a production-grade, deterministic, and auditable AI Workflow Orchestrator designed with a focus on safety, reproducibility, and deep autonomous reasoning.
Unlike traditional linear AI agents that execute tasks blindly without critical analysis, MOP introduces a Dual-Engine architecture and a Multi-Agent Debate mechanism. Before any critical decision is finalized or destructive code is executed, the agents engage in a structured discussion (debate) to identify risks, optimize solutions, and ensure maximum reliability.
graph TD
User([User Request]) --> Input[Classification & Memory Context]
Input --> Debate[Debate Engine]
subgraph Debate Engine
Analyst[Analyst Agent] --> Solution[Solution Agent]
Solution --> Critic[Critic Agent]
Critic --> Security[Security Agent]
Security --> Optimizer[Optimizer Agent]
end
Debate --> Consensus[Consensus & Aggregation]
Consensus --> Execution[Execution Layer]
Execution --> Validation{Validation Filter}
Validation -- Pass --> Store[Memory Storage & Learning]
Validation -- Fail --> Rollback[Emergency Rollback - Law 3]MOP is organized modularly, with each subsystem possessing a clearly defined responsibility:
agents/ - Agent execution and debate modes with custom prompt boundaries.debate/ - Debate flow control, argument evaluation, and aggregation algorithms.memory/ - SQLite backend for persistent memory, similarity search, and historical analysis.execution/ - Secure command and Python code execution with environment virtualization.validation/ - Pre-flight and post-flight syntax, security, and verification checks.security/ - Secrets protection, API token hygiene, and environment leak prevention.observability/ - Deep audit logging and execution tracing mapped via a unique trace_id.Built with passion and engineering discipline in the Antigravity workspace.