content image

MCPとA2A:エージェント型インターネットの静かな誕生

MCP and A2A: The Quiet Birth of the Agentic Internet

AnthropicのMCPとGoogleのA2Aは、AIエージェント時代の共通言語を巡る二大プロトコル。Linux Foundationや主要IT企業の支持を集め急速に標準化が進む今、これらは「次のインターネット」となり得るのか。
分からないところをタップすると
↓日本語訳が表示されます↓

The race to define a common language for AI agents is one of the most consequential infrastructure battles in software. Anthropic introduced the Model Context Protocol, or MCP, on November 25, 2024, as an open standard for connecting AI systems to data sources, tools, and applications. Google then announced Agent2Agent, or A2A, on April 9, 2025, as an open protocol for agents to discover one another, exchange information securely, and coordinate actions across platforms. At a technical level, they address different problems: MCP governs how an agent plugs into tools and context, while A2A governs how one agent talks to another. (anthropic.com)

What makes this moment feel historically important is not just the elegance of the protocols, but the speed of institutional alignment behind them. In December 2025, the Linux Foundation launched the Agentic AI Foundation as a neutral home for MCP, with support from major firms including AWS, Google, Microsoft, and OpenAI; the announcement said MCP already had more than 10,000 published servers. On the A2A side, the Linux Foundation said on April 9, 2026, that more than 150 organizations were supporting the protocol, with integrations across Google, Microsoft, and AWS, and the A2A community has since announced a production-ready v1.0 release. OpenAI has also built MCP directly into its developer stack, allowing models in the Responses API to use remote MCP servers and connectors as tools. (linuxfoundation.org)

So, are MCP and A2A the next internet? That is probably too grand a claim for now, but the analogy is not absurd. A2A’s own documentation now stresses a web-aligned architecture built on familiar patterns such as HTTP, JSON-RPC, streaming, and version negotiation, while official materials from both Google and the Linux Foundation repeatedly describe A2A as complementary to MCP rather than a replacement for it. In other words, the “standard war” may be less a duel than the slow emergence of a protocol stack for the agentic era. The real test will be trust: OpenAI warns that MCP servers can introduce prompt-injection and data-exfiltration risks, which means the future of agent interoperability will depend not only on openness, but also on security, governance, and careful operational discipline. (a2a-protocol.org)

by EigoBoxAI
作成:2026/06/01 07:01
レベル:超上級 (語彙目安:8000語以上)
タイプ:リーディング

まだ読んでいないコンテンツ

content image
by EigoBoxAI
作成:2026/06/01 18:05
レベル:中上級 (語彙目安:4000〜6000語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/06/01 18:03
レベル:中級 (語彙目安:2000〜2500語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/06/01 18:02
レベル:中級 (語彙目安:2000〜2500語)
タイプ:リーディング
content image
by EigoBoxAI
作成:2026/06/01 12:04
レベル:初中級 (語彙目安:1000〜2000語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/06/01 12:03
レベル:初級 (語彙目安:300〜1000語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/06/01 12:01
レベル:初中級 (語彙目安:1000〜2000語)
タイプ:リーディング
content image
by EigoBoxAI
作成:2026/06/01 07:03
レベル:超入門 (語彙目安:〜300語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/06/01 07:02
レベル:超入門 (語彙目安:〜300語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/05/31 18:04
レベル:上級 (語彙目安:6000〜8000語)
タイプ:ポッドキャスト
content image
by EigoBoxAI
作成:2026/05/31 18:02
レベル:中上級 (語彙目安:4000〜6000語)
タイプ:ポッドキャスト