The fragmented world of AI agent protocols just got significantly less fragmented.

On August 20, 2026, Google formally transferred its Agent2Agent (A2A) protocol to the Agentic AI Foundation (AAIF), a body hosted by the Linux Foundation. A2A now sits alongside Anthropic’s Model Context Protocol (MCP) under a single neutral governance structure. What started in March 2026 with 49 founding members has grown to more than 250 organisations, including AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI as Platinum signatories.

For anyone deploying AI agents in business, this is a significant structural shift, not just an industry announcement.

What A2A and MCP Actually Do

These are not competing products. They operate at different layers of the same stack, and understanding the distinction matters when you are planning an agent deployment.

MCP (Model Context Protocol) handles the vertical layer, the connection between an AI agent and the tools, data sources, and applications it needs to use. Think of it as the standard for how an agent plugs into your CRM, your databases, your calendar, or your document storage.

A2A (Agent2Agent) handles the horizontal layer, communication between AI agents themselves. When you have a customer service agent that needs to hand a task to a billing agent, or a research agent that delegates sub-tasks to specialist agents, A2A is the protocol that makes that coordination work without the agents needing to know each other’s internal logic.

Together, they cover both dimensions of what a real enterprise agent deployment needs: integration with your existing tools and coordination among multiple agents.

Why Open Governance Changes the Risk Calculation

Before this consolidation, businesses evaluating multi-agent architectures faced a legitimate concern. If Google’s A2A and Anthropic’s MCP remained proprietary protocols controlled by competing companies, building on either carried concentration risk. The standard could be changed, deprecated, or strategically restricted at any time.

Moving both to the Linux Foundation changes that calculus. The Linux Foundation model has a proven track record: Kubernetes, Node.js, and the HTTP standards all live here. Protocols under Linux Foundation governance cannot be unilaterally changed or weaponised by any single member, regardless of their size.

The growth from 49 to 250+ members in under six months is itself a signal. Standards bodies with that kind of adoption velocity tend to become permanent infrastructure. The companies building on these protocols now are building on something that will almost certainly still be here in five years.

The Vendor Lock-In Question Gets a Better Answer

One of the most common hesitations Enterprise DNA encounters when businesses are evaluating AI agent deployment is vendor lock-in. “What if we build everything on Anthropic’s tooling and they change pricing or get acquired? What if Google’s agents become the only thing that talks to Google Workspace?”

These are legitimate questions. Every enterprise software cycle has produced cautionary tales about proprietary lock-in.

With both MCP and A2A under neutral governance, the answer changes. The application layer, which vendor’s agents you use for which tasks, remains flexible. Switching from one agent framework to another no longer means rebuilding every integration from scratch. The protocols that connect agents to tools and to each other are now a stable, vendor-neutral foundation.

That does not eliminate all switching costs. Implementation experience, fine-tuned prompts, and workflow design still carry real value. But it substantially reduces the catastrophic risk scenario where a vendor decision leaves you stranded.

What This Means for Business

If you are running a wait-and-see posture on AI agents, this development is a credible signal that the infrastructure is being laid for long-term deployment. The major vendors are not competing on the protocol layer. They are competing on the application layer, which means businesses can adopt without betting on a single provider.

If you are actively deploying agents, the MCP and A2A standards are worth building to now. Workflows built on these protocols will be portable as the ecosystem matures.

If you are evaluating multi-agent workflows, the convergence of MCP and A2A under one governance body removes one of the main architectural unknowns. You can design for agent-to-agent coordination and tool integration without worrying that the underlying standards will diverge.

The businesses that have operational experience with agent-based workflows when the ecosystem reaches full interoperability will have a compounding advantage. The floor has been laid.

Enterprise DNA’s Omni Ops service deploys AI agent workforces designed around the standards that are emerging as permanent infrastructure. If you want to understand what that looks like in your specific business context, book a session with our team.