Anthropic donates MCP: the AI war is fought over protocols

What happened?

Anthropic, the creator of the Claude assistant, has transferred the rights to its Model Context Protocol (MCP) to the newly created Agentic AI Foundation under the Linux Foundation. MCP, which by December 2025 recorded nearly 100 million monthly SDK downloads and over 10,000 active servers, thus becomes an open standard governed by the community. The decision, announced alongside the foundation's creation, mirrors Google's move with Kubernetes and TensorFlow, and Google Cloud's donation of the Agent2Agent (A2A) protocol months earlier. However, the historical context reveals a deeper pattern: big tech companies often donate technologies once they have extracted their strategic value and seek to establish a standard that favors their dominant position in a higher layer.

Why is this important?

Anthropic's gesture is not altruistic. As InfoWorld notes, platform companies do not give away what they no longer care about, but what they no longer need to own because the competitive advantage has moved to another layer. By standardizing MCP, Anthropic aims for the protocol to become the lingua franca of AI agents, just as Kubernetes is for container orchestration. This benefits Anthropic because its Claude model is optimized for MCP, and the more agents use MCP, the more attractive Claude becomes as the underlying engine. Recall that Google donated Kubernetes in 2014 when container orchestration was already a priority, but its core business remained public cloud. Similarly, Anthropic donates MCP while its core business is Claude and associated services.

OpenAI, for its part, already supports remote MCP servers in its Responses API and participates in the steering committee, while Google donated A2A with backing from AWS, Cisco, Microsoft, Salesforce, SAP, and ServiceNow. The rivalry has shifted from the foundational model to the communication standard between agents. According to InfoWorld, this resembles a Game of Thrones episode where AI powers seek regime change through acts of apparent benevolence. History shows that when an open standard is driven by a dominant player, it tends to benefit that player in the long run, as happened with Android (Google) or Visual Studio Code (Microsoft).

Market consequences

This standardization will accelerate the adoption of multi-agent architectures, where different AI assistants collaborate with each other. For businesses, it means interoperability will be a non-negotiable requirement, and choosing an AI provider will also mean choosing a protocol ecosystem. Developers will gain flexibility but will also depend on standards that, while open, are driven by major players. A relevant fact: the MCP donation includes not only the protocol but also the specification and reference implementations, lowering the barrier to entry for startups and companies wanting to build compatible agents. However, the risk of fragmentation persists: if no standard prevails, we could repeat the messaging format war (like XMPP vs. proprietary protocols) that fragmented the instant messaging market.

The move recalls the browser wars or the standardization of Kubernetes: the winner is not the one with the best model, but the one who defines the common language. Anthropic, with MCP, bets that Claude will be the control plane of future agents. In terms of economic impact, consulting firm Gartner estimates that by 2027, 40% of enterprise applications will include AI agents, and interoperability will be key to avoiding vendor lock-in. This donation could accelerate that process but also consolidate the power of major cloud and model providers.

What readers should know

• It's not philanthropy: it's a platform strategy to commoditize the protocol layer and compete on the model and services layer. Anthropic follows Google's playbook with Kubernetes, TensorFlow, and Android.
• The real battlefield: no longer foundational models, but agent orchestration and communication standards. Competition has shifted up the technology stack.
• OpenAI and Google are also playing: each donates its protocols (MCP, A2A) to try to make its standard dominant. OpenAI, despite being a rival, supports MCP because it benefits from an open standard not controlled unilaterally by Anthropic.
• Impact on developers: they will have to learn and integrate these protocols but will benefit from greater interoperability. The learning curve is moderate, as MCP is designed to be simple and extensible.
• Risk of fragmentation: if no standard prevails, we could repeat the messaging format war. For now, the Linux Foundation acts as a neutral arbiter, but the influence of founding members is significant.
• Historical precedent: Google's donation of Kubernetes in 2014 led to it becoming the de facto standard for orchestration, but it also disproportionately benefited Google Cloud. Something similar could happen with Anthropic and Claude in the agent market.

"Platform companies don't give away what they no longer care about, but what they no longer need to own because the competitive advantage has moved to another layer." — InfoWorld

In summary, the MCP donation is a strategic move to position Anthropic as the center of gravity of the AI agent ecosystem, at the cost of ceding formal control of the protocol. Time will tell whether MCP becomes the dominant standard or whether fragmentation with A2A and other protocols causes confusion. What is certain is that the battle for AI is no longer just about models, but about how they connect and orchestrate.