Anemoi Agent: A2A Communication for Scalable Multi-Agent Coordination
Agent-to-agent communication server replaces context-stuffing with direct coordination, achieving 52.73% accuracy on GAIA with smaller models.
Anemoi Agent: Rethinking Multi-Agent Coordination
Anemoi challenges the dominant pattern of centralized, context-stuffed agent coordination by introducing an agent-to-agent (A2A) communication server. Rather than funneling all state and reasoning through a single planner, agents communicate directly, monitor each other’s progress, refine shared plans, and reach consensus. On the GAIA benchmark, Anemoi achieves 52.73% pass@3 accuracy using GPT-4.1-mini as planner and GPT-4o as workers—beating an OWL reproduction by +9.09 points despite using smaller models.
What changed. The largest performance gains (52%) come from collaborative refinement enabled by A2A communication, with secondary gains from reduced context redundancy (8%). Error analysis reveals that LLM capability limits (45.6%) and tooling gaps (20.6%) remain the dominant failure modes, followed by incorrect plans (11.8%) and communication latency (10.3%).
Why it matters. Multi-agent coordination is a critical cost and scalability bottleneck. Anemoi demonstrates that communication architecture—not just model size—drives both performance and efficiency, enabling smaller planner models to achieve competitive results through better information flow.
Builder takeaway. If you’re building multi-agent systems, prioritize A2A communication patterns over centralized context-stuffing. This reduces token overhead, enables collaborative refinement, and allows you to use smaller, cheaper planner models without sacrificing accuracy.