La Agentic AI Foundation (AAIF) it has become In one of the most relevant movements of the moment in the world of artificial intelligence, by trying to bring order to the growing ecosystem of Agentic AIThat is, systems capable of taking the initiative, coordinating tasks, and acting with a high degree of autonomy. Under the umbrella of the Linux FoundationThis new entity aims to become the meeting point where it is defined how agents connect with data, tools and applications in all types of environments, from the public cloud to hybrid corporate infrastructures.
The foundation starts with a significant specific weight and a combination of Big tech companies, cloud providers, and open projects which already have widespread adoption. Their roadmap involves offering neutral governance of key standards, preventing the infrastructure of the next generation of AI agents from falling into the hands of a single vendor. For Europe and Spain, where AI and data regulation is progressing well, this move simultaneously opens up a window of opportunity and a challenge to influence how these standards are designed.
What is the Agentic AI Foundation and what does it aim to achieve?
La Agentic AI Foundation It was created as a foundation under the umbrella of the Linux Foundation with a clear mission: to offer a neutral and open space to host projects, protocols, and specifications that allow AI agents to connect with enterprise tools, data, and applications in an interoperable way. The idea is that, as systems evolve from simple chatbots to agents capable of orchestrating complex workflows, there will be a common infrastructure layer that is not tied to a single company.
According to the official announcement made in the United States, AAIF is positioned as a benchmark for open standards in the field of agentic AI, with the goal of ensuring that this "critical" capability evolves with transparency and collaboration. The model is not new: the Linux Foundation is trying to replicate the approach it has followed with projects such as Linux or Kuberneteswhere a neutral foundation has served to coordinate companies with sometimes very different interests, but with a shared technological base.
A consortium dominated by cloud giants
One of the defining characteristics of the AAIF is the composition of its founding membersIn the platinum category are Amazon Web Services (AWS), Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI—a lineup that makes it clear that the major cloud and language modeling providers want a direct say in shaping the agentic infrastructure. These are joined by dozens of companies such as Oracle, SAP, Salesforce, IBM, Snowflake, and Shopify in the gold and silver categories.
This structure configures a consortium with a strong concentration of market powerFor technology managers in European companies, especially those already relying on these platforms for their workloads, the message is twofold: on the one hand, the standards promoted by the AAIF are highly likely to become de facto referencesreducing uncertainty when choosing one protocol over another. On the other hand, the fact that much of the leadership rests with the same players who dominate the cloud and generative AI fuels concerns about the extent to which the governance will be truly independent of their business priorities.
Agentic AI: from chats to coordinated autonomous agents
The launch of the Agentic AI Foundation comes at a time of phase change in AI adoptionThe industry is moving from purely conversational systems to agents capable of acting on real systemsto perform tasks, coordinate services, and collaborate with each other—like a autonomous AI assistant— with little human intervention. Jim Zemlin, executive director of the Linux Foundation, has described this transition as a leap toward autonomous agents that already rely, on a daily basis, on technologies such as Model Context Protocol (MCP), goose and the convention AGENTS.md.
The AAIF's strategy goes beyond publishing code: it's about ensuring that the next wave of intelligent automation It is based on open projects, with clear and reviewable governance rules, and not exclusively on proprietary solutions from one or two vendors. In a context where European organizations demand guarantees of security, auditing, and regulatory compliance, this approach can become a key pivot point to deploy agents in regulated sectors such as banking, energy, industry or public administration.
MCP: the protocol that aspires to be the standard "pipeline" for agents
El Model Context Protocol (MCP)Developed initially by Anthropic, it is one of the core technical pillars of the Agentic AI Foundation. It was born to solve an internal problem—how to seamlessly connect AI models with data and tools—and has rapidly evolved into a general purpose protocol to expose resources to AI agents in a standardized way.
Since Anthropic launched MCP, the protocol has gained significant traction: studies have already been published over 10.000 MCP servers covering everything from development tools to deployments in large enterprises, including Fortune 500 corporations. MCP is present on platforms such as Claude, Cursor, Microsoft Copilot, Gemini in ChromeVS Code, ChatGPT, and other popular agent and assistant environments. For engineering teams in Spain and the rest of Europe, this makes MCP a practically inevitable piece if the aim is to build upon the existing ecosystem of copilots and agents.
Mike Krieger, product manager at Anthropic, has gone so far as to describe MCP as "the industry standard for connecting AI systems with data and tools"emphasizing that its donation to the Linux Foundation seeks to preserve its open and neutral character. The underlying message is clear: if MCP becomes part of the critical infrastructure With agent AI, its control should not depend on a single company. For organizations designing their own agent platforms, MCP offers a significant tactical advantage: abstracts integration on servers that expose capabilities homogeneously, allowing workloads to be moved between different models or clouds without rewriting specific integrations, although that effective portability will depend on how each provider implements and extends the protocol.
goose: a local-first framework for agentic workflows
The second major component coming to the Agentic AI Foundation is gooseGoose is an agent framework launched by Block (parent company of Square, Cash App, Afterpay, and TIDAL). Unlike other environments focused primarily on cloud-based model calls, Goose defines itself as a system local-first that combines language models, extensible tools, and standardized integrations through MCP to reliably execute agentic workflows.
The local-first philosophy is not a minor detail for European organizations subject to strict data protection regulations. By allowing a considerable portion of the processing is kept in local or controlled environmentsGoose opens the door to architectures where only what is strictly necessary travels to the cloud, facilitating compliance with requirements of privacy, sovereignty and complianceManik Surtani, head of open source at Block, presents the initiative as a clear dilemma: either the technology that will define the coming years is encapsulated in closed products, or it is built upon open protocols and shared standards that allow a much wider range of actors to participate.
For technical teams in Spain, goose is emerging as an interesting option for prototype and deploy internal agents that connect enterprise systems such as ERP, CRM, code repositories, or data platforms, with the advantage of natively integrating the MCP stack. The big question will be whether Goose's ecosystem of plugins and extensions achieves critical mass compared to other emerging frameworks, and to what extent European companies are encouraged to actively contribute to, and not just consume, these types of open-source tools.
AGENTS.md: a minimal convention with mass adoption
The third element placed under the AAIF umbrella is AGENTS.md, a markdown convention promoted by OpenAI with a seemingly simple approach: to offer the code agents a consistent reference point, with project-specific instructions, regardless of the repository or toolchain used.
That simplicity has proven to be its greatest strength. More than 60.000 open source projects And agent frameworks have already adopted AGENTS.md, including Amp, Codex, Cursor, Devin, Factory, Gemini CLI, GitHub Copilot, Jules, and VS Code. In practice, this file functions as a kind of operational "technical specifications" For agents, this reduces ambiguity about how to interpret internal conventions of each repository and makes their behavior more predictable in complex environments, such as CI/CD pipelines or cloud platform engineering platforms.
OpenAI, which was also one of the first companies to adopt and promote MCP, has complemented AGENTS.md with tools such as ACP, Codex CLI, and the Agents and Apps SDKs, all aligned with an agentic ecosystem based on interoperable protocolsNick Cooper, a member of the OpenAI technical team, links the donation of AGENTS.md to the need to promote "open and transparent practices" that make agent development more reliable and secure. Placing this convention under the AAIF means providing it with a formal forum for evolution, with the possibility of incorporating specific requirements from regulated sectors or advanced templates, with the challenge of not losing the lightness that has driven its adoption.
Open governance as an antidote to lock-in
The Agentic AI Foundation adopts the Linux Foundation open governancewith an emphasis on interoperability and vendor neutrality. The foundation insists that its purpose is to serve as neutral guardian For MCP, goose, AGENTS.md and future related projects, inclusion criteria will be based on actual adoption, technical quality and community health, rather than financial weight or corporate sponsorship.
From the perspective of cloud customers—including many European organizations—this governance is presented as a mechanism to reduce the risk of blockage to tools and protocols tied to a single provider. Statements from executives at AWS, Google Cloud, Microsoft, Bloomberg, and Cloudflare emphasize that placing MCP and the other components under a neutral foundation allows developers and institutions invest with more confidence These standards are based on the expectation that they will remain open and community-oriented in the long term. However, experience with other standards over the past decade suggests that the real test will come when tensions arise between the commercial interests of the major players and the most beneficial technical evolution for the community.
Impact and opportunities for Spain and Europe
For Spanish companies and public administrations, the emergence of a common framework for agentic AI comes just as the European AI regulation It is becoming more established, and institutions are beginning to demand concrete guarantees regarding transparency, auditing, and data protection. The existence of a foundation like the AAIF, with the support of the Linux Foundation, offers a regulatory and technological support point so that agent projects are not built from scratch or on closed solutions that are difficult to monitor.
In sectors such as banking, energy, manufacturing, healthcare, or the public sector, the combination of MCP, goose, and AGENTS.md can facilitate the creation of advanced automation pilots that integrate legacy systems with new agent capabilities, while maintaining some degree of control over where data is processed and how agent actions are recorded. For Spain and its European partners, the key will be not just adopt not only to comply with the standards that emerge from the AAIF, but also to participate in its working groups, contribute regulated use cases and convey the requirements of the community framework on data sovereignty, explainability and systemic risk management.
The community dimension: MCP summits and next steps
The launch of the Agentic AI Foundation is accompanied by concrete actions aimed at the technical community. Obot.ai, a silver member, has donated to the foundation the MCP Dev Summit events and the associated podcast, thus consolidating a recurring meeting place for developers and companies working with the protocol. The next MCP Dev Summit will be held in New York on April 2nd and 3rd, 2026With the call for papers, registration and sponsorships already open, plans have been announced to organize a European edition also in 2026.
These types of forums will be crucial in determining whether the AAIF is capable of attract startups, technology SMEs and independent developersAnd not just to the large vendors that currently dominate the platinum and gold membership lists. The Linux Foundation has historically used conferences, working groups, and technical committees to channel feedback and turn it into roadmap decisions. The challenge now is to replicate that model in a particularly sensitive area such as that of... autonomous agentswhere concerns about safety, European regulatory compliance, labor impact and potential systemic effects converge.
With the launch of the Agentic AI Foundation, the ecosystem of the Agentic AI enters a phase of institutionalization in which protocols like MCP, frameworks like goose, and lightweight conventions like AGENTS.md could become the discrete infrastructure on which the next generations of agents in companies and public bodies will operate. The role that Europe—and, within it, countries like Spain—plays in influencing governance, contributing real-world use cases, and demanding alignment with its regulatory frameworks will be crucial for the AAIF's promise of interoperability and neutrality to translate into a more open, portable and controllable automationand not in a new layer of dependence on a few global actors.
