Study measures handshake overhead in MCP, A2A, and ACP agent protocols
Researchers released the first empirical benchmark measuring the "handshake" overhead in three major AI agent communication protocols: MCP, A2A, and ACP. The study, published on arXiv, instruments every request to track round trips, latency, wire bytes, and token costs across various scenarios, including cold-starts and degraded metadata.
Key points
- Legacy MCP handshake takes three round trips and 230 ms at 50 ms RTT.
- A2A handshake requires one round trip and 61.0 ms, significantly faster than legacy MCP.
- Metadata processing costs 24-107 tokens per capability, consuming 11% of a 200k context window for 200 capabilities.
The findings show that round-trip count is the primary driver of handshake cost. A legacy MCP handshake requires three round trips and takes 230 ms at a 50 ms RTT, whereas A2A needs only one round trip and 61.0 ms. However, token costs also matter: when an LLM processes a counterpart's metadata, the cost scales with the number of capabilities, ranging from 24 to 107 tokens per capability. For a counterpart with 200 capabilities, this consumes 11% of a 200k-token context window before the agent even begins its task.
The study notes that MCP's 2026 revision reduces its handshake from three round trips to two, and to zero on warm reconnects with era pinning. The authors have released the full testing harness, raw data, and analysis on GitHub.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- MIT historian says early AI engineers added randomness to mimic human creativity · 1 src
- Researchers introduce MemPlan for text-based planning with PDDL · 1 src
- Researchers present CiteAnchor framework for detecting hallucinated citations · 1 src
- Researchers find self‑rewarded moral argumentation makes language models decisive but pliable · 1 src
- AQL-Lite improves Arabic LLMs with curated data and partial-parameter training · 1 src
Comments
via GitHub Discussions