Blog / Standards

MCP and A2A under open governance: what it changes for companies

The two protocols that connect agents to tools and to each other no longer depend on a single vendor. That is good news for interoperability. It says nothing about what an agent is entitled to do once connected.

By Vincent Dorange, , 2 min read

9 Dec 2025

MCP joins the Linux Foundation’s Agentic AI Foundation

Two Linux Foundation announcements changed the status of agent protocols. On 23 June 2025 it launched the Agent2Agent (A2A) project, a protocol Google created in April to let agents from different vendors discover one another, exchange information and collaborate 1. On 9 December 2025 it announced the Agentic AI Foundation, home to the Model Context Protocol (MCP) contributed by Anthropic, goose contributed by Block and AGENTS.md contributed by OpenAI 2.

MCP, published by Anthropic in November 2024, connects an AI assistant to the systems where data and tools live 3. A2A lets agents talk to each other. Both now sit under neutral governance, outside the control of any single vendor.

What open governance brings

For a company, the change is concrete. A protocol held by a neutral foundation is less likely to change its rules with one supplier's strategy. Major players take part in its evolution: Amazon Web Services, Bloomberg, Cloudflare, Google and Microsoft are among the members of the Agentic AI Foundation alongside the three founding contributors 2. On the A2A side, the backers announced in June 2025 range from SAP to ServiceNow, by way of Cisco, AWS, Salesforce and Microsoft 1. The risk of being locked into a closed ecosystem shrinks accordingly.

What a protocol does not settle

A protocol says how an agent connects to a tool or to another agent. It does not say whether the agent should. An MCP server that exposes a refund function lets any technically authorised agent issue refunds; it sets no cap, no case in which a human must approve, and no person who answers for a mistake.

That distinction between authorization and authority lies at the heart of the Agentic Commerce Framework®. An agent's technical rights are granted in the information system; the power to decide is granted in writing, in its mandate. The easier protocols make connection, the more that second layer becomes necessary.

Sources

Links to the official texts and original publications.

  1. 1Linux Foundation Launches the Agent2Agent Protocol Project, Linux Foundation, 23 June 2025.
  2. 2Linux Foundation Announces the Formation of the Agentic AI Foundation, Linux Foundation, 9 December 2025.
  3. 3Introducing the Model Context Protocol, Anthropic, 25 November 2024.

Glossary terms used in this article

AI agent authorization vs authorityAI agent mandate

Vincent Dorange, creator of the Agentic Commerce Framework®

Also worth reading

95%

of organisations surveyed faced at least one agent-related incident in 12 months

Governance

95% of organisations have already had an AI agent incident, says Gartner

25

organisations interviewed; none lets its agents act with unrestricted autonomy

Governance

AI agents at work: what the OECD learned from 25 organisations

4 levers

transaction caps, budgets, approved counterparties and escalation rules, per Circle

Agentic economy

Agentic economy: agents that spend within delegated permissions

All articles