A new Agentic AI Foundation under the Linux Foundation is trying to stop AI agents from splintering into closed, incompatible ecosystems.
📌 Key Takeaways
- OpenAI, Anthropic and Block have co founded the Agentic AI Foundation (AAIF) under the Linux Foundation.
- The foundation launches with three core projects, MCP, goose and AGENTS.md, donated as open standards.
- AAIF aims to keep agentic AI interoperable, vendor neutral and governed in the open, not by one company.
- Big cloud and enterprise players are on board as members, signalling strong industry backing for shared standards.
- Developers get common “rails” for agents, cutting integration work and reducing the risk of deep vendor lock in.
Why Agentic AI Now Needs A Shared Foundation
Agentic AI is moving from toy demos to systems that take actions, call tools, and coordinate workflows on behalf of users. As that shift accelerates, every big platform is building its own agents and integration stack.
Without shared standards, those stacks drift apart. Agents risk becoming tied to one vendor’s formats, tools, and clouds, which makes interoperability hard and slows adoption for enterprises that need mixed environments, compliance controls, and clear governance around critical infrastructure.
The new Agentic AI Foundation is designed as a neutral home where those core “rails” can live outside any single product roadmap. The goal is simple, give the ecosystem a common baseline before fragmentation hardens.
Who Is Behind The Agentic AI Foundation
The foundation sits as a directed fund under the Linux Foundation, which already stewards projects like Linux, Kubernetes and PyTorch. That structure brings established processes for neutral governance, community input and long term stewardship of shared infrastructure.
AAIF is co founded by Anthropic, Block and OpenAI, with initial backing from major players including Google, Microsoft, Amazon Web Services, Cloudflare and Bloomberg. Many other companies join as platinum, gold and silver members.
This mix matters. It means the standards are shaped not only by model labs, but also by clouds, financial firms, tooling vendors and startups that will actually operate agentic systems at scale in regulated and high stake environments.
“We are seeing AI enter a new phase, as conversational systems shift to autonomous agents that can work together.” — Jim Zemlin, Executive Director, Linux Foundation
Inside MCP, goose And AGENTS.md
Model Context Protocol (MCP), created by Anthropic, is a universal, open protocol that lets AI agents connect to tools, data sources and applications in a consistent way. It already has thousands of public servers and broad adoption across major AI products.
goose, contributed by Block, is an open source, local first agent framework that uses MCP for integrations. It gives developers a reference implementation for building structured, auditable agent workflows that can run close to where data and business logic live.
AGENTS.md, contributed by OpenAI, is a simple markdown file that sits next to a repository’s README and tells coding agents how to behave, from style rules to build steps. Tens of thousands of open source projects already use it as a standard contract.
“By establishing the AAIF and contributing goose, we are taking a stand for openness so powerful tools serve everyone.” — Manik Surtani, Head of Open Source, Block
What This Means For Developers And Enterprises
For developers, these projects add up to a common language for agents. MCP defines how agents reach tools, goose offers a battle tested framework, and AGENTS.md standardises project specific guidance, so behaviours stay predictable across repos and platforms.
Enterprises get stronger guarantees that today’s investments in agentic workflows will not be stranded on one vendor’s stack. Neutral governance and open specifications make it easier to audit, self host, extend and swap components while still benefiting from ecosystem level innovation.
In the longer term, AAIF could become for agentic AI what web standards bodies were for the early internet, a place where competing companies agree on the shared rails, then differentiate on experience, performance, safety and domain expertise above that layer.
Conclusion
The Agentic AI Foundation is less about launching a shiny new product and more about locking in the boring, critical plumbing beneath AI agents. That plumbing decides whether agents stay interoperable infrastructure or fragment into closed silos.
By donating MCP, goose, and AGENTS.md into a neutral foundation at this stage, Anthropic, Block, and OpenAI are betting that open, shared standards will ultimately grow the entire market for agentic AI rather than protect narrow proprietary advantages.
📈 Latest AI News
10th December 2025
For the recent AI News, visit our site.
If you liked this article, be sure to follow us on X/Twitter and also LinkedIn for more exclusive content.
