As AI moves Chatbots and into actionable systems, the Linux foundation is launching a new group to keep incompatible, locked Product Agents locked.
The group, called the AI agent Foundation (AAIIF), will act as a neutral home for open source projects related to Ai agents. Anchoring The Aaif in Bukak is a smorgasbord of anthropopic, block, and openai.
Anthropopic is a donation Mcp (Protocol context model), a A standard way to connect Models and Agents for tools and data; Block Contribution of goose, an open-source agent framework; and carry it Agen.md On the table, a simple instruction file developers can add to the repository to tell AI Coding Translate How to behave. You can think of these tools as the basic plumbing of the era of agents.
Other members of the Aaif included AWS, Bloomberg, Cloudflare, and Google, mark the industry level to keep the locksmiths to make AI Agents reliable at scale.
At Opsiai Engineer Nick Cooper, a protocol is a shared language agent And the system works together without each developer to produce the integration from scratch.
“We need multiple (protocols) to negotiate, to communicate, and to be able to work together to deliver value to people, and communication that cannot be one provider, one company,” Cooper told TechCrunch.
Jim ZemlinThe Executive Director of the Linux Foundation, put more conversation about the launch: The goal is to future “Closed Walls”, and the orchestration is locked in a handful of platforms.
“By bringing these projects under Aaiif, we can now coordinate interoperability, safety patterns, and best practices for Ai agents,” Zemlin said.
Blokir – the fultech company behind the app and cash – is not known for its AI infrastructure, but it creates an open player with goose. AI Tech lead Brad Axen Frames is proof that Open alternatives can match property agents at scale, with thousands of engineers spending weeks on coding, data analysis, and documentation.
Many open swans open as a dual purpose for blocks.
“Let’s go out into the world and give other people a place to help themselves better,” Axen told TechCrunch. “We have a lot of contributors from open source, and all they do is add more to our company.”
Meanwhile, donating onions to the Linux Foundation Grants block access to community stress tests while positioning is an example used for Aaif building blocks – the agent framework designed for mcp and agen.md.
Anthropik made a similar move at the protocol layer, handing over MCP to the Linux Foundation. Goal: Create MCP The Neutral Infrastructure that connects AI for tools, data, and applications without one-off adapters.
“The main goal is to have enough adoption for the world to be the de facto Standard,” CO-TOGTOR David Soria Parra told TechCrunch. “We are all better off if we have an Open integration center where you can build once that is the developer and use the client.”
Contribute MCP to Aaiif Signal that the Protocol will not be controlled by a single vendor.
That governance point is central to why the Linux Foundation created the new umbrella. The organization has been leading AI infrastructure projects and developers — everything from PyTorch and Ray — but AAIF says it’s dedicated to agent standards and agent encryption, including safety and interoperability patterns.
The Aaiif structure is funded through “Directed Funds,” meaning companies can provide money through member memberships. But Zemlin of Pondok Linux argues that funding lacks the same controls: Project Roadmaps are set by a technical steering committee, and no single member disagrees with the direction.
Still, the big question is whether AAIIH is a real infrastructure or just another industry Alliance.
“An early indication of success, in addition to judging the standard, will be the development and implementation of a shared standard used by vendor agents around the world,” Zemlin said.
For Cooper Openai, success will look standard evolution: “I don’t want it to be a stagnant thing. I don’t want the protocol to be part of this foundation, and there they have to experience input.”
There are also wider ramifications: even with open government, one corporate implementation can become the default because it ships the fastest or gets the fastest usage. Zemlin says that’s not necessarily a bad thing, though. He points to open history – like kuernetes “Winning” Race Containh – as evidence of the emergence of the “dominance” of merit and vendor control. “
For developers and companies, the short-term appeal is clear: complete custom connectors, predictable agent behavior in an aware environment.
A bigger vision is more ambitious: if a tool like MCP, Agen.mdand goose become a standard infrastructure, landscape agents can switch from a closed platform to a closed software, mix-and-match from introper systems that can be used in modern systems.