The Cost of Saying Hello: Empirical Benchmark of Handshake Overhead in AI Agent Protocols (MCP, A2A, ACP)

Forum topic · 小凯 · 2026-10-24

Summary

This arXiv paper (2610.30151) presents the first empirical, phase-decomposed benchmark of handshake overhead in three major AI agent communication protocols: MCP (2025 initialize and 2026 discover eras), A2A, and ACP. Using each protocol's official SDK, the author measured round trips, per-phase latency, wire bytes, and token costs across cold-start, warm-repeat, capability-mismatch, and degraded-metadata scenarios (N=50 per cell, loopback and 50 ms emulated RTT). Key findings: round-trip count dominates handshake cost—legacy MCP requires three round trips (230 ms at 50 ms RTT) versus A2A's single round trip (61.0 ms). When an LLM reads counterpart metadata, costs scale linearly with capability inventory at 24–107 tokens per capability; a 200-capability counterpart consumes 11% of a 200k-token context window before any task work begins. MCP's 2026 revision reduced handshakes from three round trips to two, and to zero on warm reconnects via era pinning. Full harness, raw data, and analysis are open-sourced.

Paper Overview

Field: ML Author: Pranav Varshney Published: 2026-10-24 arXiv: 2610.30151

Abstract

Every interaction between AI agents begins with a handshake: the initiating agent must locate its counterpart, learn what it can do, and establish a session before the first useful message can be sent. Surveys of agent communication protocols name handshake overhead as an efficiency criterion, yet none report measurements. This paper presents the first empirical, phase-decomposed benchmark of the handshake in three major agent protocols: MCP (in both its 2025 initialize and 2026 discover eras), A2A, and ACP.

Using each protocol's official SDK, the study instruments every request for round trips, per-phase latency, wire bytes, and token cost across cold-start, warm-repeat, capability-mismatch, and degraded-metadata scenarios (N=50 per cell, on loopback and under an emulated 50 ms RTT).

Key Findings

Round-trip count dominates handshake cost. A legacy MCP handshake takes three round trips and 230 ms at 50 ms RTT; A2A needs one round trip and 61.0 ms.
Round trips are not the whole bill. When an LLM reads the counterpart's metadata, handshake cost scales with the counterpart's capability inventory at 24–107 tokens per capability.
A 200-capability counterpart spends 11% of a 200k-token context window before the agent reads its own task.
MCP's own 2026 revision cuts its handshake from three round trips to two and, with era pinning, to zero on warm reconnects—protocol evolution that corroborates the measured costs.

The full harness, raw data, and analysis are released as open source.

--- *Auto-collected on 2026-10-24*