OpenAI, Anthropic, and Block Are Teaming Up to Make AI Agents Play Nice
In a landmark move set to profoundly shape the future of artificial intelligence, industry titans OpenAI, Anthropic, and Block have announced the co-founding of the Agentic AI Foundation (AAIF). This pivotal new open-source organization, established under the esteemed umbrella of the Linux Foundation, is dedicated to forging and promoting open standards for the burgeoning field of AI agents. This collaboration signals a crucial shift in the AI landscape, moving beyond mere conversational models towards a future where intelligent agents autonomously execute tasks and interact seamlessly across digital ecosystems.
The emergence of agentic AI represents a transformative leap from the current generation of chat-based AI systems. While large language models (LLMs) have captivated the public with their ability to generate text, answer questions, and engage in conversations, AI agents take this capability several steps further. They are designed not just to understand and respond but to act on behalf of users, navigating complex digital environments, making decisions, and interacting with various services and other agents to achieve specific goals. This paradigm shift promises a potentially lucrative new era where AI agents become integral to daily life, powering a vast array of applications from personalized shopping assistants that negotiate prices and book travel, to sophisticated business agents managing transactions, customer interactions, and supply chains. The vision is one where countless AI agents routinely communicate and collaborate, making the digital world more efficient and responsive.
Recognizing the immense potential—and inherent complexities—of this agentic future, the founding companies understood that fragmented development and proprietary standards could stifle innovation and hinder widespread adoption. Just as open standards like TCP/IP and HTTP were fundamental to the internet’s explosive growth, a unified framework is essential for AI agents to achieve true interoperability. The Agentic AI Foundation aims to be that unifying force, providing a neutral ground for collaboration and standard-setting. The choice of the Linux Foundation as its host is highly strategic, leveraging its decades of experience in fostering open-source projects, offering robust legal and technological support, and ensuring transparent, community-driven governance. Beyond the three founding members, the AAIF has already garnered significant industry support, with major players like Google, Microsoft, AWS, Bloomberg, and Cloudflare signing on, underscoring the broad consensus on the need for such a collaborative effort.
A cornerstone of the AAIF’s immediate efforts involves the transfer of ownership of several widely utilized agentic technologies from the founding companies to the new foundation. This move is designed to democratize access, encourage wider adoption, and invite community contributions, accelerating the development of a robust, standardized agent ecosystem. Among these critical contributions are:
-
Anthropic’s Model Context Protocol (MCP): This protocol is vital for enabling AI agents to connect and interact effectively. In a world where agents will need to communicate across different platforms and providers, MCP provides the foundational language and structure for these interactions, ensuring clarity and compatibility. Nick Cooper, who leads work on the protocol at OpenAI, emphasized its significance, stating that while many companies already use MCP, making it an open standard will encourage a broader embrace. "That open interoperability—that open standard—really means that companies can talk across providers, and across agentic systems," Cooper explained, highlighting the critical need for seamless communication in an agent-driven world.
-
OpenAI’s Agents.md: This technology offers a structured way for programs and websites to specify rules and guidelines for coding agents. It acts as a blueprint, allowing developers to define how agents should behave, what data they can access, and what actions they can perform within specific digital environments. By open-sourcing Agents.md, OpenAI is empowering developers to create more predictable, controllable, and secure agents, fostering trust and consistency across the agent landscape.
-
Block’s Goose: Developed by Block, Goose is a powerful framework for building agents that can tap into a wide range of large language models (LLMs) to perform actions on a computer. This framework exemplifies the proactive nature of agentic AI, allowing agents to automate tasks, interact with software, and leverage the intelligence of various underlying LLMs. Manik Surtani, head of open source at Block, noted the soaring popularity of Goose over the past year. Donating Goose to the AAIF will make it significantly easier for anyone to contribute to its codebase and build upon it, further democratizing the creation of sophisticated AI agents.
These technologies, though already free to use, gain immense benefits from being housed under an open-source foundation. It shifts them from being company-specific tools to industry-wide assets, fostering a community-driven development model that promises greater transparency, stability, and innovation.
The question of openness has become increasingly crucial in the contemporary AI landscape. For many US companies, the primary monetization strategy revolves around offering access to powerful, proprietary models through application programming interfaces (APIs). However, there’s a growing debate about the long-term implications of this closed approach. Meta, for instance, previously released the weights for its highly capable Llama model, allowing anyone to download and run it, though it has recently signaled a potential shift towards a more restricted model. In stark contrast, a number of prominent Chinese AI companies, including DeepSeek, Alibaba, Moonshot AI, and Z.ai, have actively provided strong open-source models that have gained considerable traction among developers, startups, and AI researchers globally. Some experts express concern that this divergence could confer a significant strategic advantage to Chinese firms over time, as open-source models can accelerate innovation, foster a wider developer ecosystem, and potentially lead to faster real-world adoption. OpenAI itself is not entirely new to open-source efforts, with contributions like gpt-oss and Codex CLI, software for its models’ command-line interface, demonstrating a hybrid approach to development.
Srinivas Narayanan, chief technology officer of B2B applications at OpenAI, articulated the profound belief that "Open source is going to play a very big role in how AI is shaped and adopted in the real world." This sentiment underpins the AAIF’s mission. By championing open standards, OpenAI, Anthropic, and Block aim to remove barriers to the broad deployment of agentic AI across all sectors of the economy. While these standards do not mandate the use of foundation models from any specific company, the expectation is that an open ecosystem will ultimately benefit all participants, including the founding members, by expanding the overall market for agentic AI solutions.
The establishment of the AAIF under the Linux Foundation is a powerful statement about the industry’s commitment to collaborative development. Jim Zemlin, executive director of the Linux Foundation, underscored this in a statement: "MCP, Agents.md, and Goose have become essential tools for developers building this new class of agentic technologies. Bringing these projects together under the AAIF ensures they can grow with the transparency and stability that only open governance provides. The Linux Foundation is proud to serve as the neutral home where they will continue to build AI." This neutral stewardship is vital for ensuring that standards are developed equitably and serve the entire community, rather than the interests of any single commercial entity.
While open standards are inherently technologically neutral, their widespread adoption can confer significant influence upon the entities behind them. Just as organizations like ICANN and the W3C have profoundly shaped the evolution and governance of the internet, the Agentic AI Foundation holds the potential to guide how AI agents are developed and utilized on a daily basis worldwide. Should these agentic tools and standards become globally dominant, the US companies leading this initiative could wield considerable soft power, helping to determine the foundational principles and ethical considerations embedded in the fabric of future AI interactions. This long-term strategic advantage, born from collaboration and openness, highlights the profound implications of this new foundation for the global technological landscape.
In conclusion, the formation of the Agentic AI Foundation marks a critical juncture in the evolution of artificial intelligence. By fostering open standards and collaboratively developing foundational technologies, OpenAI, Anthropic, and Block, alongside a growing consortium of industry leaders, are laying the groundwork for a future where AI agents can interact seamlessly, securely, and effectively. This collaborative approach promises to unlock the full potential of agentic AI, driving innovation, accelerating adoption, and ensuring that the next wave of AI transformation benefits society as a whole, while also positioning its architects at the forefront of this new era.
