As AI moves beyond chatbots to systems that can take action, the Linux Foundation is launching a new group dedicated to preventing AI agents from splintering into a mess of incompatible, locked-down products.
The group, called the Agentic AI Foundation (AAIF), serves as a neutral hub for open source projects related to AI agents. AAIF’s launch is supported by donations from Anthropic, Block, and OpenAI.
Anthropic is contributing MCP (Model Context Protocol), a standard way to connect models and agents to tools and data. Block is contributing to its open source agent framework, Goose. OpenAI is introducing AGENTS.md, a simple instruction file that developers can add to the repository to tell AI coding tools how to behave. You can think of these tools as basic plumbing for the agent era.
Other members of the AAIF include AWS, Bloomberg, Cloudflare, and Google, demonstrating an industry-wide push for shared guardrails to ensure AI agents can be trusted at scale.
In OpenAI engineer Nick Cooper’s view, protocols are essentially shared languages that allow different agents and systems to work together without all developers having to rebuild integrations from scratch.
“Providing value to people requires multiple (protocols) to negotiate, communicate and collaborate. That kind of openness and communication is why it’s no longer one provider, one host, one company,” Cooper told TechCrunch.
Jim Zemlin, executive director of the Linux Foundation, was more blunt in a conversation at the time of the announcement: The goal is to avoid a future of “closed-wall” proprietary stacks where tool connectivity, agent behavior, and orchestration are locked behind a few platforms.
“Bringing these projects together under AAIF allows us to align interoperability, safety patterns, and best practices specifically for AI agents,” Zemlin said.
Block (the fintech company behind Square and Cash App) isn’t known for its AI infrastructure, but it’s working on openness with Goose. AI tech leader Brad Axen frames this as proof that open alternatives can compete with proprietary agents at scale, and thousands of engineers use it every week for coding, data analysis, and documentation.
Open-sourced Goose serves two purposes for Block.
“By putting this out into the world, it gives us a place where other people can help us improve,” Axen told TechCrunch. “We have a lot of contributors from open source. Everything they do to improve open source comes back to our company.”
Meanwhile, donating Goose to the Linux Foundation gives Block access to community stress testing and positions it as an example of AAIF’s vision of being an agent framework designed to plug into shared building blocks like MCP and AGENTS.md.
Anthropic is making a similar move at the protocol layer, handing MCP over to the Linux Foundation. The goal is to make MCP a neutral infrastructure that connects AI models to tools, data, and applications without endless one-off adapters.
“The main goal is to see enough adoption around the world that it becomes the de facto standard,” MCP co-creator David Soria Parra told TechCrunch. “We are all better off if we have an open integration center where as developers we can build something once and use it with any client.”
Donating MCP to AAIF indicates that the protocol is not controlled by a single vendor.
This governance point is central to why the Linux Foundation created the new umbrella in the first place. The organization already hosts major AI and developer infrastructure projects, from PyTorch and Ray to Kubernetes, but AAIF says it’s specifically aimed at agent standards and orchestration, including shared safety patterns and interoperability.
AAIF’s organization is funded through an “oriented fund,” to which companies can contribute funds through membership fees. But the Linux Foundation’s Zemlin argues that funding does not equate to stewardship. The project roadmap is set by the technical steering committee, with no single member having a unilateral say in direction.
Still, the big question is whether AAIF will become a real infrastructure or just an industry logo alliance.
“In addition to the adoption of these standards, an early indicator of success will be the development and implementation of shared standards used by vendor agents around the world,” Zemlin said.
For OpenAI’s Cooper, success will look like the evolution of standards. “I don’t want it to become stagnant. I don’t want these protocols to become part of this foundation. And that’s where they’ve been sitting for two years. The protocols need to evolve and continually accept further input.”
There are also more subtle results. Even with open governance, one company’s implementation may become the default simply because it shipped first or was used the most. But Zemlin says that’s not necessarily a bad thing. He points to the history of open source, such as Kubernetes “winning” the container race, as evidence that “advantage comes from merit, not vendor control.”
The short-term appeal for developers and businesses is clear. It takes less time to build custom connectors, agent behavior is more predictable across your codebase, and deployment in security-sensitive environments is easier.
Bigger visions are more ambitious. As tools like MCP, AGENTS.md, and Goose become standard infrastructure, the agent landscape could move from a closed platform to a world of open, combinable software reminiscent of the interoperable systems that built the modern web.
