Scaling enterprise and agentic AI will rely heavily on industry players agreeing on standards. From model context protocol (MCP) to Agent 2 Agent (A2A) and beyond, interoperability protocols give organizations a common language and way to identify agents. This could significantly fast track adoption of multiple agents across industries.

The Linux Foundation, Anthropic, OpenAI, Block, and other companies in the AI space created the Agentic AI Foundation to help further standardize the evolution of open-source AI. This brings a large chunk of interoperability and agentic standards under one roof, potentially streamlining the development of the different protocols. The move would significantly streamline how enterprise AI builders contribute to projects; ensure no single vendor controls a company's roadmap; guarantee broad interoperability; and offer faster time to deployment. For enterprise AI builders, it marks greater confidence that the agents they create can work with other agents without fear of a vendor removing access to the protocol on a whim.

As part of AAIF’s creation, Anthropic will bring its widely adopted MCP to the Linux Foundation, and OpenAI contributed its AGENTS.md. Block principal engineer Brad Axen told VentureBeat that his company brought its AI agent framework, goose, to AAIF, likening it to a reference implementation of agents. 

Other members of the AAIF include AWS, Bloomberg, Cloudflare, Google, Microsoft, IBM, Datado, Oracle, JetBrains, Salesforce, SAP, Snowflake, Hugging Face, Uber, and Zapier, among others. 

Agentic protocols like MCP, AGNTCY, and A2A help create a way for agents to talk to each other securely, whereas AGENTS.md gives coding agents access to context. Even before these standards were donated to the Linux Foundation, enterprises could build with them and sometimes contribute in their evolution. But ownership of the standards remained with the companies that created them, and they can steer the standard or remove investment in it if the wanted to. No matter how widespread MCP is, Anthropic still owned and controlled it.

But by giving MCP to a neutral third party like the Linux Foundation, the community gets more of a say on how to develop the standards, and it scales in a way that benefits everyone.

How does this help enterprise AI builders

With more enterprises exploring agents, AI builders want to ensure that the tools and platforms they create work across different ecosystems. So they build applications with the standards in mind, hoping that the protocols are not changed too much to benefit their corporate owners.

Opening up the governance of these projects removes the fear of vendor lock-in, where expansion and decisions about the protocol could be changed at the whim of its owner. Enterprise AI developers can be assured that when they build toward a standard, they can expect shared APIs and find repeatable deployment patterns. 

“Protocols enable a world where people's context is shared, and let the systems that they work with be more powerful,” Axen said. “And MCP is a big part of that, so customers of Block can build a system that is watching your email to say like, ‘Oh,  someone has emailed me requesting an appointment. I'm just going to book that for you. And that's the kind of connectivity that something like MCP enables. So even for corporations, this means we can build solutions that help our customers more by connecting different parts of the ecosystem.”

Open-sourcing standards 

Several companies have “donated” their interoperability standards to the Linux Foundation, effectively ceding control over the project.

Anthropic said in a blog post that MCP’s governance model will remain unchanged, with the project’s maintainers continuing to “prioritize community input and transparent decision-making.”  

“Bringing these and future projects under the AAIF will foster innovation across the agentic AI ecosystem and ensure these foundational technologies remain neutral, open, and community-driven,” Anthropic said. 

Google donated A2A, its interoperability layer, to the foundation in June, arguing in a press release that the Linux Foundation “will ensure that this critical component remains vendor-agnostic and community-driven. This move is designed to accelerate the adoption and development of the A2A protocol by providing a robust framework for open collaboration, intellectual property management and long-term stewardship.”

Cisco also donated AGNTCY to the foundation this year. OpenAI said in its blog post that bringing AGENTs.md to a larger group highlights the importance of open standards to “make agents safe, easier to build and more portable across tools and platforms.”

And it’s not just interoperability standards; the Linux Foundation is also home to the machine learning framework PyTorch, making it a hub for many tools used to build AI systems. 

Block’s Axen told VentureBeat that having more standards under one roof shows how far the AI industry has gone.

“It shows that the industry is maturing around open source, and we understand how powerful it is,” Axen said. “It gives all of these protocols a durable home to survive in the long run.”