Tom Hundley🤖 Ghostwritten by Claude Opus 4.5 · Curated by Tom Hundley
This article was written by Claude Opus 4.5 and curated for publication by Tom Hundley.
The AI industry just had its TCP/IP moment.
On December 9, 2025, something remarkable happened: OpenAI, Anthropic, and Block—companies that compete fiercely for AI market dominance—set aside their rivalries to launch the Agentic AI Foundation (AAIF) under the Linux Foundation. They weren't alone. Amazon Web Services, Google, Microsoft, Bloomberg, and Cloudflare joined as platinum members, with IBM, Salesforce, Oracle, and dozens of others signing on as gold and silver members.
Why would competitors collaborate so publicly? Because they all recognized the same threat: without shared standards, the AI agent ecosystem was heading toward the same fragmented mess that plagued early computing, mobile development, and IoT. And that fragmentation would hurt everyone—vendors, enterprises, and developers alike.
Here's the challenge: AI agents are no longer just chatbots that answer questions. They're autonomous systems that take initiative, make decisions, and execute complex multi-step workflows with minimal human direction. They browse the web, write code, manage databases, schedule meetings, and coordinate with other agents to complete sophisticated tasks.
But without standards, every AI vendor was building their own proprietary ecosystem. An agent built for Claude couldn't easily connect to tools designed for ChatGPT. A workflow automation created in one platform couldn't be ported to another. Enterprise customers were being locked into single-vendor stacks, and developers were drowning in integration work.
The AAIF was created to solve this—to be, as many have described it, "the W3C of agentic AI" or "the USB-C of AI integration."
The foundation launched with three critical open-source projects that address different layers of the agentic AI stack:
Donated by Anthropic, the Model Context Protocol has rapidly become the standard for connecting AI models to external tools, data sources, and applications. Think of MCP as a universal adapter—instead of building custom integrations for every service your AI needs to access, you build one MCP connection and gain access to thousands of compatible tools.
The numbers tell the story of MCP's adoption:
Mike Krieger, Anthropic's Chief Product Officer, explained the decision to donate MCP: "What started as a solution to our own internal problems has become the industry standard for connecting AI systems. By donating MCP to the Linux Foundation, we're ensuring it remains vendor-neutral and community-driven—following the same model that made Kubernetes and Node.js successful."
Contributed by OpenAI, AGENTS.md is deceptively simple but remarkably powerful. It's a markdown file format that gives AI coding agents consistent, project-specific instructions for working in any codebase.
Think of how a README.md file helps human developers understand a project. AGENTS.md serves the same purpose for AI agents—telling them how to navigate the codebase, which commands to run, what coding standards to follow, and how to handle pull requests.
Since its release in August 2025, AGENTS.md has been adopted by more than 60,000 open-source projects and is supported by virtually every major AI coding tool: GitHub Copilot, Cursor, Devin, Gemini CLI, VS Code, and OpenAI's own Codex.
Nick Cooper from OpenAI emphasized why standardization matters: "Open, transparent practices make agent development more predictable and safe. When every repository follows the same conventions for agent instructions, AI tools become more reliable across the entire ecosystem."
Block (formerly Square) contributed Goose, an open-source AI agent framework that serves as a reference implementation showing how MCP and agentic AI should work together in practice.
Goose isn't just a demo—it's a production-ready agent framework that can build entire projects from scratch, execute code, debug failures, orchestrate complex workflows, and interact with external APIs autonomously. It runs locally on your machine (local-first architecture), giving developers full control over their agent's behavior and data.
Manik Surtani, Block's Open Source Lead, framed the stakes clearly: "Agentic AI can remain closed, siloed, and controlled by a handful of companies—or it can be driven by open standards that benefit everyone. We chose the latter."
The AAIF's membership roster reads like a who's-who of enterprise technology:
Platinum Members (the governing tier):
Gold Members (17 organizations including):
Silver Members: 20+ additional organizations including Apify, Chronosphere, Elasticsearch, and Hugging Face
This broad membership ensures that standards developed by the AAIF will reflect the needs of the entire industry—not just the AI model providers, but also the cloud platforms, enterprise software vendors, developer tool makers, and infrastructure providers that make up the agentic AI ecosystem.
The timing of the AAIF's formation isn't coincidental. Enterprise AI adoption has reached an inflection point:
The market is exploding: The agentic AI market is projected to grow from $7.55 billion in 2025 to $199 billion by 2034—a 43.84% compound annual growth rate. 85% of enterprises expect to implement AI agents by end of 2025.
Adoption is already widespread: According to recent surveys, 79% of organizations report some level of agentic AI adoption, with 96% planning to expand usage in 2025. Over 60% of new enterprise AI deployments now include agentic capabilities.
ROI is compelling: Organizations deploying agentic AI report an average projected ROI of 171%, with U.S. enterprises specifically forecasting 192% returns. Two-thirds of senior executives say their agentic AI initiatives are already delivering measurable productivity or business value.
But here's the catch: these benefits only materialize fully when agents can work together across systems. An agent that can only access tools from one vendor, or that can't coordinate with agents running on different platforms, delivers a fraction of its potential value.
The AAIF's standards solve this by ensuring:
The AAIF explicitly models itself on successful open infrastructure initiatives. As Block stated in their announcement: "The Internet, Linux, and the Web succeeded precisely because they were open."
Jim Zemlin, Executive Director of the Linux Foundation, positioned the AAIF within this tradition: "We are seeing AI enter a new phase, as conversational systems shift to autonomous agents that can work together. Within just one year, MCP, AGENTS.md, and Goose have become essential tools for developers building this new class of agentic technologies."
The governance structure mirrors other successful Linux Foundation projects like Kubernetes and PyTorch:
The AAIF has already announced its first major community event: the MCP Dev Summit North America, scheduled for April 2-3, 2026 in New York City. A European summit is planned for later in 2026.
Beyond events, the foundation will focus on:
The Agentic AI Foundation represents a pivotal moment in AI's evolution. Just as HTML, TCP/IP, and USB-C enabled explosive innovation by providing shared standards, the AAIF's protocols—MCP, AGENTS.md, and the patterns exemplified by Goose—provide the foundation for an interoperable agentic AI ecosystem.
For enterprises, this means:
For developers, this means:
The age of proprietary AI silos is ending. The age of interoperable AI agents has begun.
The Agentic AI Foundation is hosted by the Linux Foundation. For more information, visit aaif.io or the Linux Foundation announcement.
Discover more content:
