Anemoi: A Semi-Centralized Multi-agent Systems Based on Agent-to-Agent Communication MCP server from Coral Protocol

2025-08-23

AI Summary

Existing general-purpose multi-agent systems (MAS) rely on a centralized planner and unidirectional prompt chaining, leading to excessive dependence on large language models (LLMs), inefficient inter-agent communication, information redundancy, and difficulty in identifying bottlenecks. This paper proposes a semi-centralized MAS architecture built upon the Coral protocol, featuring an Agent-to-Agent communication server compliant with the Model Context Protocol (MCP). It enables structured message passing, runtime state sharing, and dynamic collaboration while diminishing the role of the central planner—thereby supporting real-time progress monitoring, bottleneck detection, and adaptive plan optimization. Evaluated on the GAIA benchmark using GPT-4.1-mini as the planner, our approach achieves 52.73% accuracy, outperforming the strongest open-source baseline (OWL) by 9.09%. The core contributions are a decentralized collaboration mechanism and a lightweight, LLM-friendly communication paradigm, significantly improving performance, scalability, and cost-efficiency.

Technology Category

• Multiagent Systems: Multiagent Planning
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

Recent advances in generalist multi-agent systems (MAS) have largely followed a context-engineering plus centralized paradigm, where a planner agent coordinates multiple worker agents through unidirectional prompt passing. While effective under strong planner models, this design suffers from two critical limitations: (1) strong dependency on the planner's capability, which leads to degraded performance when a smaller LLM powers the planner; and (2) limited inter-agent communication, where collaboration relies on costly prompt concatenation and context injection, introducing redundancy and information loss. To address these challenges, we propose Anemoi, a semi-centralized MAS built on the Agent-to-Agent (A2A) communication MCP server from Coral Protocol. Unlike traditional designs, Anemoi enables structured and direct inter-agent collaboration, allowing all agents to monitor progress, assess results, identify bottlenecks, and propose refinements in real time. This paradigm reduces reliance on a single planner, supports adaptive plan updates, and minimizes redundant context passing, resulting in more scalable and cost-efficient execution. Evaluated on the GAIA benchmark, Anemoi achieved 52.73% accuracy with a small LLM (GPT-4.1-mini) as the planner, surpassing the strongest open-source baseline OWL (43.63%) by +9.09% under identical LLM settings. Our implementation is publicly available at https://github.com/Coral-Protocol/Anemoi.

Problem

• Reducing planner dependency in multi-agent systems
• Enabling direct inter-agent communication and collaboration
• Minimizing redundant context passing and information loss

Innovation

• Semi-centralized multi-agent system design
• Agent-to-Agent communication MCP server
• Real-time structured inter-agent collaboration monitoring