OpenAI, Anthropic, and Block Launch Agentic AI Foundation

OpenAI, Anthropic, and Block have officially co-founded the Agentic AI Foundation (AAIF), a new organization operating under the Linux Foundation. This alliance aims to establish open technical standards and interoperability protocols for the next generation of AI agents in the enterprise sector.

The landscape of AI development faces a significant hurdle: fragmentation. Without unified protocols, AI agents developed by different providers cannot communicate or execute complex workflows across disparate platforms effectively. The establishment of the AAIF aims to address this friction by creating a neutral environment for collaboration.

Nick Cooper, Member of Technical Staff, OpenAI, says that transforming proprietary tools into open standards is a strategic move to encourage broad adoption among developers and corporations. "That open interoperability, that open standard, really means that companies can speak between vendors and between AI systems," says Cooper.

The Shift to Transactional AI

The focus is shifting from passive Large Language Models (LLMs) — primarily used for text generation and chat interfaces — toward "agentic AI." Unlike their predecessors, agentic systems are designed to execute actions autonomously. They can navigate the web, negotiate commercial processes, and manage complex B2B workflows with minimal human oversight.

Srinivas Narayanan, Head of Engineering for Applied AI, OpenAI, envisions a business environment where high volumes of AI agents communicate routinely to execute transactions. For this vision to materialize, the industry requires a shared infrastructure, he says. Narayanan adds that open source will play a major role in the configuration and adoption of AI in the real world. By aligning on open standards, the founding members aim to replicate the success of early web protocols, ensuring that automated interactions occur seamlessly across the digital economy.

Technical Contributions and Governance

To jumpstart the foundation, the three founding companies have transferred the ownership of key technologies to the AAIF. This ensures these tools remain neutral and open for community contribution:

• Model Context Protocol (MCP): Originally developed by Anthropic, this standard functions as a universal connectivity layer, allowing assistants and agents to connect with data sources and other tools.

• Agents.md: A contribution from OpenAI, this serves as a repository that allows programs and websites to specify rules for how agents should interact with their content and code.

• Goose: A framework developed by Block designed for building agents that execute actions directly on computer interfaces.

Manik Surtani, Head of Open Source, Block, says that the Goose agent exploits a wide range of LLMs to perform actions on a computer and has increased in popularity in the last year. He emphasizes that donating Goose to the new AAIF makes it easier for anyone to contribute to the code base and build upon it.

The AAIF operates under the Linux Foundation, which provides the legal and technological framework for high-level open-source projects."MCP, Agents.md, and Goose have become essential tools for developers building this new class of agentic technologies," says Jim Zemlin, Executive Director, The Linux Foundation. "Bringing these projects under the AAIF ensures they can grow with the transparency and stability that only open governance provides."

Industry Adoption and Strategic Implications

The initiative has garnered immediate support from major infrastructure and service providers. In addition to the founders, members include Google, Microsoft, AWS, Bloomberg, and Cloudflare. This consolidation suggests a unified effort by US technology leaders to define the architectural rules of the global AI market.

This move toward open standards also holds geopolitical significance. While US companies typically monetize access through closed models and APIs (with exceptions such as the Llama models from Meta), competitors in China have gained traction with powerful open-source models. Companies such as DeepSeek, Alibaba, Moonshot AI, and Z.ai offer tools that are popular among startups and researchers.

The AAIF allows US corporations to counter this trend by controlling the standards of interconnection rather than just the models themselves. Similar to how the ICANN and the W3C determined the evolution of the web, the AAIF has the potential to define how agentic AI operates globally.