A group of leading U.S. artificial intelligence firms has formed a new foundation to establish open standards for “agentic” AI. The founding members, OpenAI, Anthropic, and Block, have pooled their proprietary agent- and AI-related technologies into a new open-source project called the Agentic AI Foundation (AAIF), under the auspices of the Linux Foundation.
This development follows tensions in the global race for dominance in artificial intelligence, leading U.S. AI firms and policymakers to unite around a new push to preserve American primacy.
Open Standards Drive Innovation and Cross-Platform Collaboration
Cloudflare CTO Dane Knecht noted that open standards and protocols are critical for establishing an evolving developer ecosystem for building agents. He added, “They ensure anyone can build agents across platforms without the fear of vendor lock-in.”
American companies face a dilemma because they are seeking continuous income from closed APIs, even as they are falling behind in fundamental AI development, risking long-term irrelevance to China. This means American companies must standardize their approach for MCP and agentic AI, allowing them to focus on building better models rather than being locked into an ecosystem.
The foundation establishes both a practical partnership and a milestone for community open-sourcing, with adversaries uniting around a single goal of standardization rather than fragmentation. It also makes open-source development easier and more accessible for users worldwide, including those in China.
Anthropic donated its Model Context Protocol (MCP), a library that allows AIs to utilize tools creatively outside API calls, to the Linux Foundation. Since its introduction a year ago, MCP has gained traction, with over 10,000 active servers, best-in-class support from platforms including ChatGPT, Gemini, Microsoft Copilot, and VS Code, as well as 97 million monthly SDK downloads.
“Open-source software is key to creating a world with secure and innovative AI tools for agentic applications,” Anthropic wrote.
OpenAI added AGENTS.md, used in 60,000 repositories as a lightweight specification to standardize AI agent instructions. Block contributed Goose, a local-first agent platform. All three now operate under the neutral governance of the Linux Foundation.
Foundation Strengthens U.S. AI Leadership Amid Rising Competition from China
The timing is ideal for U.S. AI leadership, as Chinese companies have released more open models, while American firms have focused on closed APIs. Jim Zemlin, the Linux Foundation executive director, added: “We are seeing AI come into a new kind of phase — that of conversational systems that will transition to autonomous agents that can co-exist.”
The foundation addresses a fundamental weakness: relying on Chinese open-source models reduces the need for large U.S. cloud providers and APIs. Platinum members include Amazon, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI.
Some of the gold group members include Cisco, Datadog, Docker, IBM, Oracle, SAP, Snowflake, and Twilio. Other silver members include Hugging Face, Uber, and SUSE, among others. A single company does not head it.
China’s competitive advantage stems from a concerted policy. Rather than giant AI plants, a focus on low-cost adaptability and modular innovation enables Chinese firms to create open-source weights that developers can use to build upon their technology.
The U.S. recognizes the stakes. The Trump administration’s AI Action Plan highlights the geostrategic value of open-source and open-weight models, noting they could become global standards in business and academic research worldwide.