Skip to main content

Documentation Overview

Welcome to the Multi-Agent Lifecycle Protocol (MPLP) documentation. MPLP is the open standard for defining, coordinating, and executing multi-agent systems with full observability and governance.

This site serves as the normative specification for protocol implementers and the guide for agent developers.

🚀 Getting Started

If you are new to MPLP, start here:

📚 Core Concepts

The protocol is divided into several key areas:

Architecture

Understanding the L1–L4 Layered Architecture:

  • L1 Core: State management and identity.
  • L2 Coordination: Governance and consensus.
  • L3 Execution: Runtime and orchestration.
  • L4 Integration: Infrastructure and transport.

Modules

The composable building blocks of an agent:

  • Memory: Short-term, long-term, and episodic memory standards.
  • Planning: Reasoning and task decomposition.
  • Tools: Standardized tool interfaces.
  • User Interaction: Human-in-the-loop patterns.

Profiles

Standardized capabilities for different agent types (e.g., conversational-basic, autonomous-planner).

Observability

The "Black Box" recorder for agents. Learn about the OpenTelemetry-based tracing standards and event taxonomy.

🛠️ For Developers

  • SDKs: Official libraries for TypeScript and Python.
  • Tests: Validate your implementation against the Golden Test Suite.

🤝 Contributing

MPLP is an open standard. We welcome contributions to the specification and the standard library.