- The Agentic AI Foundation is created under the umbrella of the Linux Foundation as a neutral home for open agentic AI.
- MCP, Goose and AGENTS.md are consolidated as foundational standards for connecting agents with data, tools and projects.
- Major technology companies such as AWS, Google, Microsoft, OpenAI, Anthropic, and Block support the initiative and fund its development.
- The goal is to avoid closed ecosystems and promote interoperability, security, and community governance in AI agents.
The call Agentic AI, that in which systems no longer just answer questions but make decisions and chain tasks together on their own, It is entering a phase of organization and shared rules.. In this context The Agentic AI Foundation (AAIF) has been launched, a new coordinated effort under the umbrella of the Linux Foundation that It aims to establish the technical and governance foundations for these intelligent agents..
The plan involves bringing together in a single neutral space several projects that are already functioning as “Basic plumbing” of the agent era: The Model Context Protocol (MCP) from Anthropic, the framework Goose developed by Block and the specification AGENTS.md Driven by OpenAI, these initiatives aim to provide companies, European public administrations, and independent developers with a common infrastructure on which to build more interoperable AI solutions that are less dependent on a single vendor.
A new foundation to bring order to the era of AI agents
La AAIF was created as a fund managed within the Linux Foundation, co-founded by anthropicBlock and OpenAIand backed by companies like Amazon Web Services (AWS), Google, Microsoft, Cloudflare and BloombergIts stated mission is that agentic AI evolves in a transparent, collaborative, and public interest-oriented mannerthrough strategic investment in open projects and the definition of shared technical standards.
From the Linux Foundation, its executive director, Jim Zemlin, has insisted that The priority is to prevent a future dominated by “walled gardens” of proprietary technologyWhere tool connectivity, agent behavior, and workflow orchestration are locked within a few platforms. Integrate MCP, Goose, and AGENTS.md under a single umbrella This would allow for better coordination of interoperability, security models, and best practices. specific to agents.
The financial structure of the The foundation is based on a system of memberships and fees. that feed that directed fundHowever, it is emphasized that financial contributions do not translate into absolute control: the technical roadmap will depend on steering committees formed by the community and not on a single company, replicating models that have already worked in projects such as Linux, Kubernetes or PyTorch.
For Europe and Spain, where EU institutions have been debating for months how to fit advanced AI into the framework of European AI ActThe existence of such a foundation under an agent as recognized as the Linux Foundation provides a more solid basis for promoting pilots and public projects on truly open infrastructures.
MCP: a “USB-C” for connecting AI models with data and tools
El Model Context Protocol (MCP) It is probably the most mature piece of the set. Anthropic presented it as an open and universal standard for linking AI applications with external systems, and has described it on more than one occasion as a kind of “USB-C of the AI world”: a single connector that allows you to integrate models with databases, corporate APIs or cloud services without having to develop custom integrations for each case.
According to figures shared by Anthropic and the Linux Foundation itself, there are already more than 10.000 public MCP serversencompassing everything from development tools to internal deployments in large companies. Platforms as well-known as ClaudeChatGPT, Gemini, Microsoft Copilot, Cursor, and Visual Studio Code have incorporated support for this protocol, which has made it easier for AI agents to work with very diverse data sources and utilities.
In terms of infrastructure, providers such as AWS, Google Cloud, Microsoft Azure or Cloudflare They offer specific deployment and operational mechanisms for MCP, enabling European companies to build compatible agent architectures in virtually any relevant cloud environment. This widespread availability reinforces the idea that MCP acts as a de facto standard to connect models and tools.
The protocol not only defines how connections are made, but also key design aspects such as asynchronous operations, server identity, official extensions, and the most reliable behavior. “stateless” possible, something especially relevant when dealing with sensitive or regulated data, such as financial or health data in the European context.
Goose: a local-first agent framework for complex workflows
Along with MCP, the foundation hosts GooseGoose is an open-source agent framework developed by Block, the parent company of services like Square and Cash App. It was designed from the outset as a platform. local-firstThat is, prepared so that a large part of the processing can be carried out in environments controlled by the organization itself, and not exclusively in the cloud.
The framework combines language models, extensible tools and native MCP integration to build structured agentic workflowsIn practice, this allows for the chaining of autonomous tasks related to programming, data analysis, or document management, while maintaining clear rules of security and regulatory compliance.
According to Block officials, Thousands of engineers use Goose regularly. for internal development and analysis tasks, which has served as a large-scale testing ground. By releasing it as open-source software and donating it to the Linux Foundation, Block pursues a dual purpose: to benefit from contributions from the global community and to make Goose a very tangible example of how a framework aligned with AAIF standards should work.
For European companies subject to strict regulations—from PSD2 in the financial sector to GDPR regarding personal data—Goose's local-first philosophy fits reasonably well with the requirements of control over where and how data is processed, something that is not always easy with completely closed solutions.
AGENTS.md: Clear and consistent instructions for code agents
The third technical pillar of the Agentic AI Foundation is AGENTS.mdA simple, text-based specification that OpenAI introduced as a standard way to describe how agents should behave in a given software project. Instead of relying on scattered documentation or implicit conventions, this file offers coding tools a unique and legible reference point.
The proposal has been remarkably well-received: it is estimated that more than 60.000 projects and agent frameworks They have already incorporated AGENTS.md, including popular developer tools such as Cursor, Devin, GitHub Copilot, Amp, Gemini CLI, and VS Code. This widespread integration makes agent behavior more predictable, reducing friction when working with diverse repositories and different toolchains.
OpenAI, in addition to promoting this convention, has been one of the first participants in the MCP ecosystem, contributing utilities such as SDKs, CLIs, and kits for applications built on these standards. The company emphasizes that The protocols operate as a common language This prevents each developer from having to reinvent their own integrations, making it easier for different agents and systems to cooperate without the need for ad hoc bilateral agreements.
In a European scenario where large corporations, SMEs, and public administrations coexist, using very different tools and programming languages, a simple and standardized mechanism like AGENTS.md can make all the difference. reliable agents and unpredictable agents that require too much human supervision.
Broad industry alliance and role of the Linux Foundation
The Agentic AI Foundation launches with a membership list that includes many of the most influential names in the global technology sector. Companies such as [list of companies would go here] are among its highest-level supporters. AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAIThese are joined by dozens of additional players between the Gold and Silver levels, including companies in cloud infrastructure, payments, development, observability and enterprise software.
Among the partners are names also well-known in the European technology sector, such as Cisco, IBM, Oracle, Salesforce, SAP, Snowflake, Hugging Face, SUSE or EricssonThe presence of these types of actors is relevant for Spain and the EU because many of them already collaborate on standardization initiatives and regulated sectors at the community level, which facilitates the integration of these new protocols into projects financed with European funds or sponsored by public bodies.
The Linux Foundation, for its part, brings decades of experience managing critical open-source projects. It has been home to technologies that underpin much of today's digital infrastructure, from the linux kernel From Kubernetes to Node.js, OpenSSF, PyTorch, and RISC-V, this track record is one of the arguments used to reinforce the idea that the AAIF will not be a mere alliance of logos, but a real working environment with proven processes of collaboration and neutral governance.
In any case, within the sector itself it is acknowledged that the true measure of success will be to see if The agents that deploy suppliers and companies end up using these standards massively.Or, conversely, will the foundation remain more declarative than operational? Some officials suggest that a key indicator will be the emergence of shared specifications adopted by major agent vendors, such as a common API standard for conversational or orchestration tools.
Interoperability, security, and concern for risks
The creation of the Agentic AI Foundation comes at a time when organizations, both in Europe and in other regions, are incorporating AI agents in business processes at a rapid pace. Industry reports place the percentage of companies experimenting with agents or in the pilot deployment phase at around two-thirds, with a majority of executives willing to increase spending in this area in the coming years.
These advances are accompanied by clear concerns: nearly all IT and security managers consulted in various studies express concern about operational and cybersecurity risks These issues are related to the autonomy of the agents, especially when they operate on critical systems or handle sensitive information. There is a lack of solid, shared guidelines on how these systems should be configured, audited, and monitored.
The AAIF is conceived precisely as an antidote to a possible ecosystem fragmentationwhere each provider uses its own protocol, authentication mechanisms, and permission models. Without minimum consensus—for example, on how OAuth access is managed in agent contexts or how agent actions should be tracked for auditing purposes—the risk is ending up with a landscape full of technological islands that are difficult to interoper.
Among the open debates in the developer community is the possibility that the foundation will help define shared interfaces similar to what was once the standardization of web APIs or data formats, for example browsers with agentive navigationThe idea is that, if the main providers are already drawing inspiration from similar patterns for their agent services, it makes sense to converge towards a common specification that also includes compatibility testbeds.
At the same time, there is no shortage of critical voices warning of the difficulty of maintaining protocols and tools in the long term. Some developers question whether technologies like MCP will maintain their dominant position or whether more efficient alternatives will emerge that could displace them. In any case, the Linux Foundation often points out that, in the world of free software, hegemony usually arises from... technical meritocracy rather than commercial impositions, citing the example of Kubernetes in the field of containers.
Given this scenario, the Agentic AI Foundation is emerging as a new meeting point for those who want the next wave of intelligent automation to be based on open protocols, neutral governance, and true interoperabilityRather than relying on closed and difficult-to-combine solutions, its evolution over the coming years will serve to measure the extent to which the sector is capable of agreeing on common rules for a technology that, by all accounts, will be central to the European digital economy.
I am a technology enthusiast who has turned his "geek" interests into a profession. I have spent more than 10 years of my life using cutting-edge technology and tinkering with all kinds of programs out of pure curiosity. Now I have specialized in computer technology and video games. This is because for more than 5 years I have been writing for various websites on technology and video games, creating articles that seek to give you the information you need in a language that is understandable to everyone.
If you have any questions, my knowledge ranges from everything related to the Windows operating system as well as Android for mobile phones. And my commitment is to you, I am always willing to spend a few minutes and help you resolve any questions you may have in this internet world.