According to Wired, OpenAI, Anthropic, and Block are forming a new group called the Agentic AI Foundation under the Linux Foundation. The three founding companies are transferring ownership of key agent-building technologies to this new body, including Anthropic’s Model Context Protocol (MCP), OpenAI’s Agents.md specification, and Block’s Goose framework. Other major backers who have already signed on include Google, Microsoft, AWS, Bloomberg, and Cloudflare. The goal is to establish open standards so that AI agents from different companies and systems can communicate and work together. This reflects a broader industry shift from simple chat-based AI to more autonomous “agentic” systems that can take actions on a user’s behalf.
The Open Standard Play
Here’s the thing: these technologies, like Anthropic’s MCP and OpenAI’s Agents.md, were already free to use. So what’s the big deal? It’s all about governance and perceived neutrality. By moving them to a foundation—especially one under the well-respected Linux Foundation—they’re trying to signal that this isn’t just an OpenAI or Anthropic project. It’s a communal resource. Nick Cooper from OpenAI hit the nail on the head: that “open interoperability” means companies can build systems that talk across providers. They’re trying to prevent the exact kind of walled-garden, fragmented mess that has plagued other tech sectors. It’s a preemptive strike against incompatibility.
Why Agentic AI Is The Next Big Bet
Everyone’s talking about agents now. But what does that actually mean? Basically, we’re moving beyond an AI that just answers your question. The next phase is AI that *does* things for you—books a flight, negotiates a price with another AI, manages your supply chain. That’s the “potentially lucrative new paradigm” they’re all chasing. Think about it. If you’re building complex industrial systems that rely on AI for monitoring and control, you need those AI components to communicate flawlessly with hardware and other software. This is where robust, standardized agent protocols become critical infrastructure, not just nice-to-have software. For businesses integrating automation, the reliability of these connections is everything.
The Broader Openness War
This foundation move is one front in a much larger battle over openness in AI. On one side, you have the US heavyweights (OpenAI, Anthropic) who primarily make money through closed, proprietary API access. On the other, you have a surge of strong open-source models, notably from Chinese firms like Moonshot AI and Z.ai. Meta’s Llama played in the open space but is getting more closed. So there’s a real strategic tension. Is the future a handful of closed, centralized AI gods? Or a sprawling ecosystem of open, interoperable agents? This foundation seems like an attempt by the API-money crowd to have it both ways: keep their core models proprietary but open up the *plumbing* between agents. It’s a clever way to foster an ecosystem they still ultimately control.
Will It Actually Work?
So, will this alliance make AI agents play nice? It’s a good first step. Having Google, Microsoft, and AWS at the table is huge—it means the major cloud platforms are theoretically on board. But let’s be skeptical. These companies are still ferocious competitors. Standards bodies can move slowly, and commercial incentives to create “special” optimizations that lock users in are powerful. The proof will be in the adoption. If developers start building on Goose and MCP because they trust the foundation, and big applications emerge, then it wins. If one company decides to push its own proprietary agent protocol harder, the whole thing could splinter. The foundation is betting that the fear of a fragmented, useless agent future is scarier than the urge to dominate. We’ll see who’s right.
